title
stringlengths 7
246
| abstract
stringlengths 6
3.31k
|
|---|---|
Algorithms for Lipschitz Learning on Graphs | We develop fast algorithms for solving regression problems on graphs where
one is given the value of a function at some vertices, and must find its
smoothest possible extension to all vertices. The extension we compute is the
absolutely minimal Lipschitz extension, and is the limit for large $p$ of
$p$-Laplacian regularization. We present an algorithm that computes a minimal
Lipschitz extension in expected linear time, and an algorithm that computes an
absolutely minimal Lipschitz extension in expected time $\widetilde{O} (m n)$.
The latter algorithm has variants that seem to run much faster in practice.
These extensions are particularly amenable to regularization: we can perform
$l_{0}$-regularization on the given values in polynomial time and
$l_{1}$-regularization on the initial function values and on graph edge weights
in time $\widetilde{O} (m^{3/2})$.
|
Monotonous (Semi-)Nonnegative Matrix Factorization | Nonnegative matrix factorization (NMF) factorizes a non-negative matrix into
product of two non-negative matrices, namely a signal matrix and a mixing
matrix. NMF suffers from the scale and ordering ambiguities. Often, the source
signals can be monotonous in nature. For example, in source separation problem,
the source signals can be monotonously increasing or decreasing while the
mixing matrix can have nonnegative entries. NMF methods may not be effective
for such cases as it suffers from the ordering ambiguity. This paper proposes
an approach to incorporate notion of monotonicity in NMF, labeled as monotonous
NMF. An algorithm based on alternating least-squares is proposed for recovering
monotonous signals from a data matrix. Further, the assumption on mixing matrix
is relaxed to extend monotonous NMF for data matrix with real numbers as
entries. The approach is illustrated using synthetic noisy data. The results
obtained by monotonous NMF are compared with standard NMF algorithms in the
literature, and it is shown that monotonous NMF estimates source signals well
in comparison to standard NMF algorithms when the underlying sources signals
are monotonous.
|
Multi-Object Classification and Unsupervised Scene Understanding Using
Deep Learning Features and Latent Tree Probabilistic Models | Deep learning has shown state-of-art classification performance on datasets
such as ImageNet, which contain a single object in each image. However,
multi-object classification is far more challenging. We present a unified
framework which leverages the strengths of multiple machine learning methods,
viz deep learning, probabilistic models and kernel methods to obtain
state-of-art performance on Microsoft COCO, consisting of non-iconic images. We
incorporate contextual information in natural images through a conditional
latent tree probabilistic model (CLTM), where the object co-occurrences are
conditioned on the extracted fc7 features from pre-trained Imagenet CNN as
input. We learn the CLTM tree structure using conditional pairwise
probabilities for object co-occurrences, estimated through kernel methods, and
we learn its node and edge potentials by training a new 3-layer neural network,
which takes fc7 features as input. Object classification is carried out via
inference on the learnt conditional tree model, and we obtain significant gain
in precision-recall and F-measures on MS-COCO, especially for difficult object
categories. Moreover, the latent variables in the CLTM capture scene
information: the images with top activations for a latent node have common
themes such as being a grasslands or a food scene, and on on. In addition, we
show that a simple k-means clustering of the inferred latent nodes alone
significantly improves scene classification performance on the MIT-Indoor
dataset, without the need for any retraining, and without using scene labels
during training. Thus, we present a unified framework for multi-object
classification and unsupervised scene understanding.
|
Deconstructing Principal Component Analysis Using a Data Reconciliation
Perspective | Data reconciliation (DR) and Principal Component Analysis (PCA) are two
popular data analysis techniques in process industries. Data reconciliation is
used to obtain accurate and consistent estimates of variables and parameters
from erroneous measurements. PCA is primarily used as a method for reducing the
dimensionality of high dimensional data and as a preprocessing technique for
denoising measurements. These techniques have been developed and deployed
independently of each other. The primary purpose of this article is to
elucidate the close relationship between these two seemingly disparate
techniques. This leads to a unified framework for applying PCA and DR. Further,
we show how the two techniques can be deployed together in a collaborative and
consistent manner to process data. The framework has been extended to deal with
partially measured systems and to incorporate partial knowledge available about
the process model.
|
Using PCA to Efficiently Represent State Spaces | Reinforcement learning algorithms need to deal with the exponential growth of
states and actions when exploring optimal control in high-dimensional spaces.
This is known as the curse of dimensionality. By projecting the agent's state
onto a low-dimensional manifold, we can represent the state space in a smaller
and more efficient representation. By using this representation during
learning, the agent can converge to a good policy much faster. We test this
approach in the Mario Benchmarking Domain. When using dimensionality reduction
in Mario, learning converges much faster to a good policy. But, there is a
critical convergence-performance trade-off. By projecting onto a
low-dimensional manifold, we are ignoring important data. In this paper, we
explore this trade-off of convergence and performance. We find that learning in
as few as 4 dimensions (instead of 9), we can improve performance past learning
in the full dimensional space at a faster convergence rate.
|
Can deep learning help you find the perfect match? | Is he/she my type or not? The answer to this question depends on the personal
preferences of the one asking it. The individual process of obtaining a full
answer may generally be difficult and time consuming, but often an approximate
answer can be obtained simply by looking at a photo of the potential match.
Such approximate answers based on visual cues can be produced in a fraction of
a second, a phenomenon that has led to a series of recently successful dating
apps in which users rate others positively or negatively using primarily a
single photo. In this paper we explore using convolutional networks to create a
model of an individual's personal preferences based on rated photos. This
introduced task is difficult due to the large number of variations in profile
pictures and the noise in attractiveness labels. Toward this task we collect a
dataset comprised of $9364$ pictures and binary labels for each. We compare
performance of convolutional models trained in three ways: first directly on
the collected dataset, second with features transferred from a network trained
to predict gender, and third with features transferred from a network trained
on ImageNet. Our findings show that ImageNet features transfer best, producing
a model that attains $68.1\%$ accuracy on the test set and is moderately
successful at predicting matches.
|
Making Sense of Hidden Layer Information in Deep Networks by Learning
Hierarchical Targets | This paper proposes an architecture for deep neural networks with hidden
layer branches that learn targets of lower hierarchy than final layer targets.
The branches provide a channel for enforcing useful information in hidden layer
which helps in attaining better accuracy, both for the final layer and hidden
layers. The shared layers modify their weights using the gradients of all cost
functions higher than the branching layer. This model provides a flexible
inference system with many levels of targets which is modular and can be used
efficiently in situations requiring different levels of results according to
complexity. This paper applies the idea to a text classification task on 20
Newsgroups data set with two level of hierarchical targets and a comparison is
made with training without the use of hidden layer branches.
|
Highway Networks | There is plenty of theoretical and empirical evidence that depth of neural
networks is a crucial ingredient for their success. However, network training
becomes more difficult with increasing depth and training of very deep networks
remains an open problem. In this extended abstract, we introduce a new
architecture designed to ease gradient-based training of very deep networks. We
refer to networks with this architecture as highway networks, since they allow
unimpeded information flow across several layers on "information highways". The
architecture is characterized by the use of gating units which learn to
regulate the flow of information through a network. Highway networks with
hundreds of layers can be trained directly using stochastic gradient descent
and with a variety of activation functions, opening up the possibility of
studying extremely deep and efficient architectures.
|
Order-Revealing Encryption and the Hardness of Private Learning | An order-revealing encryption scheme gives a public procedure by which two
ciphertexts can be compared to reveal the ordering of their underlying
plaintexts. We show how to use order-revealing encryption to separate
computationally efficient PAC learning from efficient $(\epsilon,
\delta)$-differentially private PAC learning. That is, we construct a concept
class that is efficiently PAC learnable, but for which every efficient learner
fails to be differentially private. This answers a question of Kasiviswanathan
et al. (FOCS '08, SIAM J. Comput. '11).
To prove our result, we give a generic transformation from an order-revealing
encryption scheme into one with strongly correct comparison, which enables the
consistent comparison of ciphertexts that are not obtained as the valid
encryption of any message. We believe this construction may be of independent
interest.
|
Block Basis Factorization for Scalable Kernel Matrix Evaluation | Kernel methods are widespread in machine learning; however, they are limited
by the quadratic complexity of the construction, application, and storage of
kernel matrices. Low-rank matrix approximation algorithms are widely used to
address this problem and reduce the arithmetic and storage cost. However, we
observed that for some datasets with wide intra-class variability, the optimal
kernel parameter for smaller classes yields a matrix that is less well
approximated by low-rank methods. In this paper, we propose an efficient
structured low-rank approximation method -- the Block Basis Factorization (BBF)
-- and its fast construction algorithm to approximate radial basis function
(RBF) kernel matrices. Our approach has linear memory cost and floating-point
operations for many machine learning kernels. BBF works for a wide range of
kernel bandwidth parameters and extends the domain of applicability of low-rank
approximation methods significantly. Our empirical results demonstrate the
stability and superiority over the state-of-art kernel approximation
algorithms.
|
Visualization of Tradeoff in Evaluation: from Precision-Recall & PN to
LIFT, ROC & BIRD | Evaluation often aims to reduce the correctness or error characteristics of a
system down to a single number, but that always involves trade-offs. Another
way of dealing with this is to quote two numbers, such as Recall and Precision,
or Sensitivity and Specificity. But it can also be useful to see more than
this, and a graphical approach can explore sensitivity to cost, prevalence,
bias, noise, parameters and hyper-parameters.
Moreover, most techniques are implicitly based on two balanced classes, and
our ability to visualize graphically is intrinsically two dimensional, but we
often want to visualize in a multiclass context. We review the dichotomous
approaches relating to Precision, Recall, and ROC as well as the related LIFT
chart, exploring how they handle unbalanced and multiclass data, and deriving
new probabilistic and information theoretic variants of LIFT that help deal
with the issues associated with the handling of multiple and unbalanced
classes.
|
Optimal Time-Series Motifs | Motifs are the most repetitive/frequent patterns of a time-series. The
discovery of motifs is crucial for practitioners in order to understand and
interpret the phenomena occurring in sequential data. Currently, motifs are
searched among series sub-sequences, aiming at selecting the most frequently
occurring ones. Search-based methods, which try out series sub-sequence as
motif candidates, are currently believed to be the best methods in finding the
most frequent patterns.
However, this paper proposes an entirely new perspective in finding motifs.
We demonstrate that searching is non-optimal since the domain of motifs is
restricted, and instead we propose a principled optimization approach able to
find optimal motifs. We treat the occurrence frequency as a function and
time-series motifs as its parameters, therefore we \textit{learn} the optimal
motifs that maximize the frequency function. In contrast to searching, our
method is able to discover the most repetitive patterns (hence optimal), even
in cases where they do not explicitly occur as sub-sequences. Experiments on
several real-life time-series datasets show that the motifs found by our method
are highly more frequent than the ones found through searching, for exactly the
same distance threshold.
|
Kernel Spectral Clustering and applications | In this chapter we review the main literature related to kernel spectral
clustering (KSC), an approach to clustering cast within a kernel-based
optimization setting. KSC represents a least-squares support vector machine
based formulation of spectral clustering described by a weighted kernel PCA
objective. Just as in the classifier case, the binary clustering model is
expressed by a hyperplane in a high dimensional space induced by a kernel. In
addition, the multi-way clustering can be obtained by combining a set of binary
decision functions via an Error Correcting Output Codes (ECOC) encoding scheme.
Because of its model-based nature, the KSC method encompasses three main steps:
training, validation, testing. In the validation stage model selection is
performed to obtain tuning parameters, like the number of clusters present in
the data. This is a major advantage compared to classical spectral clustering
where the determination of the clustering parameters is unclear and relies on
heuristics. Once a KSC model is trained on a small subset of the entire data,
it is able to generalize well to unseen test points. Beyond the basic
formulation, sparse KSC algorithms based on the Incomplete Cholesky
Decomposition (ICD) and $L_0$, $L_1, L_0 + L_1$, Group Lasso regularization are
reviewed. In that respect, we show how it is possible to handle large scale
data. Also, two possible ways to perform hierarchical clustering and a soft
clustering method are presented. Finally, real-world applications such as image
segmentation, power load time-series clustering, document clustering and big
data learning are considered.
|
Risk Bounds For Mode Clustering | Density mode clustering is a nonparametric clustering method. The clusters
are the basins of attraction of the modes of a density estimator. We study the
risk of mode-based clustering. We show that the clustering risk over the
cluster cores --- the regions where the density is high --- is very small even
in high dimensions. And under a low noise condition, the overall cluster risk
is small even beyond the cores, in high dimensions.
|
Reinforcement Learning Neural Turing Machines - Revised | The Neural Turing Machine (NTM) is more expressive than all previously
considered models because of its external memory. It can be viewed as a broader
effort to use abstract external Interfaces and to learn a parametric model that
interacts with them.
The capabilities of a model can be extended by providing it with proper
Interfaces that interact with the world. These external Interfaces include
memory, a database, a search engine, or a piece of software such as a theorem
verifier. Some of these Interfaces are provided by the developers of the model.
However, many important existing Interfaces, such as databases and search
engines, are discrete.
We examine feasibility of learning models to interact with discrete
Interfaces. We investigate the following discrete Interfaces: a memory Tape, an
input Tape, and an output Tape. We use a Reinforcement Learning algorithm to
train a neural network that interacts with such Interfaces to solve simple
algorithmic tasks. Our Interfaces are expressive enough to make our model
Turing complete.
|
On Regret-Optimal Learning in Decentralized Multi-player Multi-armed
Bandits | We consider the problem of learning in single-player and multiplayer
multiarmed bandit models. Bandit problems are classes of online learning
problems that capture exploration versus exploitation tradeoffs. In a
multiarmed bandit model, players can pick among many arms, and each play of an
arm generates an i.i.d. reward from an unknown distribution. The objective is
to design a policy that maximizes the expected reward over a time horizon for a
single player setting and the sum of expected rewards for the multiplayer
setting. In the multiplayer setting, arms may give different rewards to
different players. There is no separate channel for coordination among the
players. Any attempt at communication is costly and adds to regret. We propose
two decentralizable policies, $\tt E^3$ ($\tt E$-$\tt cubed$) and $\tt
E^3$-$\tt TS$, that can be used in both single player and multiplayer settings.
These policies are shown to yield expected regret that grows at most as
O($\log^{1+\epsilon} T$). It is well known that $\log T$ is the lower bound on
the rate of growth of regret even in a centralized case. The proposed
algorithms improve on prior work where regret grew at O($\log^2 T$). More
fundamentally, these policies address the question of additional cost incurred
in decentralized online learning, suggesting that there is at most an
$\epsilon$-factor cost in terms of order of regret. This solves a problem of
relevance in many domains and had been open for a while.
|
fastFM: A Library for Factorization Machines | Factorization Machines (FM) are only used in a narrow range of applications
and are not part of the standard toolbox of machine learning models. This is a
pity, because even though FMs are recognized as being very successful for
recommender system type applications they are a general model to deal with
sparse and high dimensional features. Our Factorization Machine implementation
provides easy access to many solvers and supports regression, classification
and ranking tasks. Such an implementation simplifies the use of FM's for a wide
field of applications. This implementation has the potential to improve our
understanding of the FM model and drive new development.
|
Optimal Learning via the Fourier Transform for Sums of Independent
Integer Random Variables | We study the structure and learnability of sums of independent integer random
variables (SIIRVs). For $k \in \mathbb{Z}_{+}$, a $k$-SIIRV of order $n \in
\mathbb{Z}_{+}$ is the probability distribution of the sum of $n$ independent
random variables each supported on $\{0, 1, \dots, k-1\}$. We denote by ${\cal
S}_{n,k}$ the set of all $k$-SIIRVs of order $n$.
In this paper, we tightly characterize the sample and computational
complexity of learning $k$-SIIRVs. More precisely, we design a computationally
efficient algorithm that uses $\widetilde{O}(k/\epsilon^2)$ samples, and learns
an arbitrary $k$-SIIRV within error $\epsilon,$ in total variation distance.
Moreover, we show that the {\em optimal} sample complexity of this learning
problem is $\Theta((k/\epsilon^2)\sqrt{\log(1/\epsilon)}).$ Our algorithm
proceeds by learning the Fourier transform of the target $k$-SIIRV in its
effective support. Its correctness relies on the {\em approximate sparsity} of
the Fourier transform of $k$-SIIRVs -- a structural property that we establish,
roughly stating that the Fourier transform of $k$-SIIRVs has small magnitude
outside a small set.
Along the way we prove several new structural results about $k$-SIIRVs. As
one of our main structural contributions, we give an efficient algorithm to
construct a sparse {\em proper} $\epsilon$-cover for ${\cal S}_{n,k},$ in total
variation distance. We also obtain a novel geometric characterization of the
space of $k$-SIIRVs. Our characterization allows us to prove a tight lower
bound on the size of $\epsilon$-covers for ${\cal S}_{n,k}$, and is the key
ingredient in our tight sample complexity lower bound.
Our approach of exploiting the sparsity of the Fourier transform in
distribution learning is general, and has recently found additional
applications.
|
Interleaved Text/Image Deep Mining on a Large-Scale Radiology Database
for Automated Image Interpretation | Despite tremendous progress in computer vision, there has not been an attempt
for machine learning on very large-scale medical image databases. We present an
interleaved text/image deep learning system to extract and mine the semantic
interactions of radiology images and reports from a national research
hospital's Picture Archiving and Communication System. With natural language
processing, we mine a collection of representative ~216K two-dimensional key
images selected by clinicians for diagnostic reference, and match the images
with their descriptions in an automated manner. Our system interleaves between
unsupervised learning and supervised learning on document- and sentence-level
text collections, to generate semantic labels and to predict them given an
image. Given an image of a patient scan, semantic topics in radiology levels
are predicted, and associated key-words are generated. Also, a number of
frequent disease types are detected as present or absent, to provide more
specific interpretation of a patient scan. This shows the potential of
large-scale learning and prediction in electronic patient records available in
most modern clinical institutions.
|
Self-Expressive Decompositions for Matrix Approximation and Clustering | Data-aware methods for dimensionality reduction and matrix decomposition aim
to find low-dimensional structure in a collection of data. Classical approaches
discover such structure by learning a basis that can efficiently express the
collection. Recently, "self expression", the idea of using a small subset of
data vectors to represent the full collection, has been developed as an
alternative to learning. Here, we introduce a scalable method for computing
sparse SElf-Expressive Decompositions (SEED). SEED is a greedy method that
constructs a basis by sequentially selecting incoherent vectors from the
dataset. After forming a basis from a subset of vectors in the dataset, SEED
then computes a sparse representation of the dataset with respect to this
basis. We develop sufficient conditions under which SEED exactly represents low
rank matrices and vectors sampled from a unions of independent subspaces. We
show how SEED can be used in applications ranging from matrix approximation and
denoising to clustering, and apply it to numerous real-world datasets. Our
results demonstrate that SEED is an attractive low-complexity alternative to
other sparse matrix factorization approaches such as sparse PCA and
self-expressive methods for clustering.
|
A novel plasticity rule can explain the development of sensorimotor
intelligence | Grounding autonomous behavior in the nervous system is a fundamental
challenge for neuroscience. In particular, the self-organized behavioral
development provides more questions than answers. Are there special functional
units for curiosity, motivation, and creativity? This paper argues that these
features can be grounded in synaptic plasticity itself, without requiring any
higher level constructs. We propose differential extrinsic plasticity (DEP) as
a new synaptic rule for self-learning systems and apply it to a number of
complex robotic systems as a test case. Without specifying any purpose or goal,
seemingly purposeful and adaptive behavior is developed, displaying a certain
level of sensorimotor intelligence. These surprising results require no system
specific modifications of the DEP rule but arise rather from the underlying
mechanism of spontaneous symmetry breaking due to the tight
brain-body-environment coupling. The new synaptic rule is biologically
plausible and it would be an interesting target for a neurobiolocal
investigation. We also argue that this neuronal mechanism may have been a
catalyst in natural evolution.
|
Empirical Evaluation of Rectified Activations in Convolutional Network | In this paper we investigate the performance of different types of rectified
activation functions in convolutional neural network: standard rectified linear
unit (ReLU), leaky rectified linear unit (Leaky ReLU), parametric rectified
linear unit (PReLU) and a new randomized leaky rectified linear units (RReLU).
We evaluate these activation function on standard image classification task.
Our experiments suggest that incorporating a non-zero slope for negative part
in rectified activation units could consistently improve the results. Thus our
findings are negative on the common belief that sparsity is the key of good
performance in ReLU. Moreover, on small scale dataset, using deterministic
negative slope or learning it are both prone to overfitting. They are not as
effective as using their randomized counterpart. By using RReLU, we achieved
75.68\% accuracy on CIFAR-100 test set without multiple test or ensemble.
|
Large-scale Classification of Fine-Art Paintings: Learning The Right
Metric on The Right Feature | In the past few years, the number of fine-art collections that are digitized
and publicly available has been growing rapidly. With the availability of such
large collections of digitized artworks comes the need to develop multimedia
systems to archive and retrieve this pool of data. Measuring the visual
similarity between artistic items is an essential step for such multimedia
systems, which can benefit more high-level multimedia tasks. In order to model
this similarity between paintings, we should extract the appropriate visual
features for paintings and find out the best approach to learn the similarity
metric based on these features. We investigate a comprehensive list of visual
features and metric learning approaches to learn an optimized similarity
measure between paintings. We develop a machine that is able to make
aesthetic-related semantic-level judgments, such as predicting a painting's
style, genre, and artist, as well as providing similarity measures optimized
based on the knowledge available in the domain of art historical
interpretation. Our experiments show the value of using this similarity measure
for the aforementioned prediction tasks.
|
An $O(n\log(n))$ Algorithm for Projecting Onto the Ordered Weighted
$\ell_1$ Norm Ball | The ordered weighted $\ell_1$ (OWL) norm is a newly developed generalization
of the Octogonal Shrinkage and Clustering Algorithm for Regression (OSCAR)
norm. This norm has desirable statistical properties and can be used to perform
simultaneous clustering and regression. In this paper, we show how to compute
the projection of an $n$-dimensional vector onto the OWL norm ball in
$O(n\log(n))$ operations. In addition, we illustrate the performance of our
algorithm on a synthetic regression test.
|
Reinforced Decision Trees | In order to speed-up classification models when facing a large number of
categories, one usual approach consists in organizing the categories in a
particular structure, this structure being then used as a way to speed-up the
prediction computation. This is for example the case when using
error-correcting codes or even hierarchies of categories. But in the majority
of approaches, this structure is chosen \textit{by hand}, or during a
preliminary step, and not integrated in the learning process. We propose a new
model called Reinforced Decision Tree which simultaneously learns how to
organize categories in a tree structure and how to classify any input based on
this structure. This approach keeps the advantages of existing techniques (low
inference complexity) but allows one to build efficient classifiers in one
learning step. The learning algorithm is inspired by reinforcement learning and
policy-gradient techniques which allows us to integrate the two steps (building
the tree, and learning the classifier) in one single algorithm.
|
The Configurable SAT Solver Challenge (CSSC) | It is well known that different solution strategies work well for different
types of instances of hard combinatorial problems. As a consequence, most
solvers for the propositional satisfiability problem (SAT) expose parameters
that allow them to be customized to a particular family of instances. In the
international SAT competition series, these parameters are ignored: solvers are
run using a single default parameter setting (supplied by the authors) for all
benchmark instances in a given track. While this competition format rewards
solvers with robust default settings, it does not reflect the situation faced
by a practitioner who only cares about performance on one particular
application and can invest some time into tuning solver parameters for this
application. The new Configurable SAT Solver Competition (CSSC) compares
solvers in this latter setting, scoring each solver by the performance it
achieved after a fully automated configuration step. This article describes the
CSSC in more detail, and reports the results obtained in its two instantiations
so far, CSSC 2013 and 2014.
|
A Comprehensive Study On The Applications Of Machine Learning For
Diagnosis Of Cancer | Collectively, lung cancer, breast cancer and melanoma was diagnosed in over
535,340 people out of which, 209,400 deaths were reported [13]. It is estimated
that over 600,000 people will be diagnosed with these forms of cancer in 2015.
Most of the deaths from lung cancer, breast cancer and melanoma result due to
late detection. All of these cancers, if detected early, are 100% curable. In
this study, we develop and evaluate algorithms to diagnose Breast cancer,
Melanoma, and Lung cancer. In the first part of the study, we employed a
normalised Gradient Descent and an Artificial Neural Network to diagnose breast
cancer with an overall accuracy of 91% and 95% respectively. In the second part
of the study, an artificial neural network coupled with image processing and
analysis algorithms was employed to achieve an overall accuracy of 93% A naive
mobile based application that allowed people to take diagnostic tests on their
phones was developed. Finally, a Support Vector Machine algorithm incorporating
image processing and image analysis algorithms was developed to diagnose lung
cancer with an accuracy of 94%. All of the aforementioned systems had very low
false positive and false negative rates. We are developing an online network
that incorporates all of these systems and allows people to collaborate
globally.
|
Re-scale boosting for regression and classification | Boosting is a learning scheme that combines weak prediction rules to produce
a strong composite estimator, with the underlying intuition that one can obtain
accurate prediction rules by combining "rough" ones. Although boosting is
proved to be consistent and overfitting-resistant, its numerical convergence
rate is relatively slow. The aim of this paper is to develop a new boosting
strategy, called the re-scale boosting (RBoosting), to accelerate the numerical
convergence rate and, consequently, improve the learning performance of
boosting. Our studies show that RBoosting possesses the almost optimal
numerical convergence rate in the sense that, up to a logarithmic factor, it
can reach the minimax nonlinear approximation rate. We then use RBoosting to
tackle both the classification and regression problems, and deduce a tight
generalization error estimate. The theoretical and experimental results show
that RBoosting outperforms boosting in terms of generalization.
|
Fast Differentially Private Matrix Factorization | Differentially private collaborative filtering is a challenging task, both in
terms of accuracy and speed. We present a simple algorithm that is provably
differentially private, while offering good performance, using a novel
connection of differential privacy to Bayesian posterior sampling via
Stochastic Gradient Langevin Dynamics. Due to its simplicity the algorithm
lends itself to efficient implementation. By careful systems design and by
exploiting the power law behavior of the data to maximize CPU cache bandwidth
we are able to generate 1024 dimensional models at a rate of 8.5 million
recommendations per second on a single PC.
|
Estimation from Pairwise Comparisons: Sharp Minimax Bounds with Topology
Dependence | Data in the form of pairwise comparisons arises in many domains, including
preference elicitation, sporting competitions, and peer grading among others.
We consider parametric ordinal models for such pairwise comparison data
involving a latent vector $w^* \in \mathbb{R}^d$ that represents the
"qualities" of the $d$ items being compared; this class of models includes the
two most widely used parametric models--the Bradley-Terry-Luce (BTL) and the
Thurstone models. Working within a standard minimax framework, we provide tight
upper and lower bounds on the optimal error in estimating the quality score
vector $w^*$ under this class of models. The bounds depend on the topology of
the comparison graph induced by the subset of pairs being compared via its
Laplacian spectrum. Thus, in settings where the subset of pairs may be chosen,
our results provide principled guidelines for making this choice. Finally, we
compare these error rates to those under cardinal measurement models and show
that the error rates in the ordinal and cardinal settings have identical
scalings apart from constant pre-factors.
|
A Fixed-Size Encoding Method for Variable-Length Sequences with its
Application to Neural Network Language Models | In this paper, we propose the new fixed-size ordinally-forgetting encoding
(FOFE) method, which can almost uniquely encode any variable-length sequence of
words into a fixed-size representation. FOFE can model the word order in a
sequence using a simple ordinally-forgetting mechanism according to the
positions of words. In this work, we have applied FOFE to feedforward neural
network language models (FNN-LMs). Experimental results have shown that without
using any recurrent feedbacks, FOFE based FNN-LMs can significantly outperform
not only the standard fixed-input FNN-LMs but also the popular RNN-LMs.
|
Learning and Optimization with Submodular Functions | In many naturally occurring optimization problems one needs to ensure that
the definition of the optimization problem lends itself to solutions that are
tractable to compute. In cases where exact solutions cannot be computed
tractably, it is beneficial to have strong guarantees on the tractable
approximate solutions. In order operate under these criterion most optimization
problems are cast under the umbrella of convexity or submodularity. In this
report we will study design and optimization over a common class of functions
called submodular functions. Set functions, and specifically submodular set
functions, characterize a wide variety of naturally occurring optimization
problems, and the property of submodularity of set functions has deep
theoretical consequences with wide ranging applications. Informally, the
property of submodularity of set functions concerns the intuitive "principle of
diminishing returns. This property states that adding an element to a smaller
set has more value than adding it to a larger set. Common examples of
submodular monotone functions are entropies, concave functions of cardinality,
and matroid rank functions; non-monotone examples include graph cuts, network
flows, and mutual information.
In this paper we will review the formal definition of submodularity; the
optimization of submodular functions, both maximization and minimization; and
finally discuss some applications in relation to learning and reasoning using
submodular functions.
|
Blind Compressive Sensing Framework for Collaborative Filtering | Existing works based on latent factor models have focused on representing the
rating matrix as a product of user and item latent factor matrices, both being
dense. Latent (factor) vectors define the degree to which a trait is possessed
by an item or the affinity of user towards that trait. A dense user matrix is a
reasonable assumption as each user will like/dislike a trait to certain extent.
However, any item will possess only a few of the attributes and never all.
Hence, the item matrix should ideally have a sparse structure rather than a
dense one as formulated in earlier works. Therefore we propose to factor the
ratings matrix into a dense user matrix and a sparse item matrix which leads us
to the Blind Compressed Sensing (BCS) framework. We derive an efficient
algorithm for solving the BCS problem based on Majorization Minimization (MM)
technique. Our proposed approach is able to achieve significantly higher
accuracy and shorter run times as compared to existing approaches.
|
Context-Aware Mobility Management in HetNets: A Reinforcement Learning
Approach | The use of small cell deployments in heterogeneous network (HetNet)
environments is expected to be a key feature of 4G networks and beyond, and
essential for providing higher user throughput and cell-edge coverage. However,
due to different coverage sizes of macro and pico base stations (BSs), such a
paradigm shift introduces additional requirements and challenges in dense
networks. Among these challenges is the handover performance of user equipment
(UEs), which will be impacted especially when high velocity UEs traverse
picocells. In this paper, we propose a coordination-based and context-aware
mobility management (MM) procedure for small cell networks using tools from
reinforcement learning. Here, macro and pico BSs jointly learn their long-term
traffic loads and optimal cell range expansion, and schedule their UEs based on
their velocities and historical rates (exchanged among tiers). The proposed
approach is shown to not only outperform the classical MM in terms of UE
throughput, but also to enable better fairness. In average, a gain of up to
80\% is achieved for UE throughput, while the handover failure probability is
reduced up to a factor of three by the proposed learning based MM approaches.
|
A Survey of Predictive Modelling under Imbalanced Distributions | Many real world data mining applications involve obtaining predictive models
using data sets with strongly imbalanced distributions of the target variable.
Frequently, the least common values of this target variable are associated with
events that are highly relevant for end users (e.g. fraud detection, unusual
returns on stock markets, anticipation of catastrophes, etc.). Moreover, the
events may have different costs and benefits, which when associated with the
rarity of some of them on the available training data creates serious problems
to predictive modelling techniques. This paper presents a survey of existing
techniques for handling these important applications of predictive analytics.
Although most of the existing work addresses classification tasks (nominal
target variables), we also describe methods designed to handle similar problems
within regression tasks (numeric target variables). In this survey we discuss
the main challenges raised by imbalanced distributions, describe the main
approaches to these problems, propose a taxonomy of these methods and refer to
some related problems within predictive modelling.
|
Integrating K-means with Quadratic Programming Feature Selection | Several data mining problems are characterized by data in high dimensions.
One of the popular ways to reduce the dimensionality of the data is to perform
feature selection, i.e, select a subset of relevant and non-redundant features.
Recently, Quadratic Programming Feature Selection (QPFS) has been proposed
which formulates the feature selection problem as a quadratic program. It has
been shown to outperform many of the existing feature selection methods for a
variety of applications. Though, better than many existing approaches, the
running time complexity of QPFS is cubic in the number of features, which can
be quite computationally expensive even for moderately sized datasets. In this
paper we propose a novel method for feature selection by integrating k-means
clustering with QPFS. The basic variant of our approach runs k-means to bring
down the number of features which need to be passed on to QPFS. We then enhance
this idea, wherein we gradually refine the feature space from a very coarse
clustering to a fine-grained one, by interleaving steps of QPFS with k-means
clustering. Every step of QPFS helps in identifying the clusters of irrelevant
features (which can then be thrown away), whereas every step of k-means further
refines the clusters which are potentially relevant. We show that our iterative
refinement of clusters is guaranteed to converge. We provide bounds on the
number of distance computations involved in the k-means algorithm. Further,
each QPFS run is now cubic in number of clusters, which can be much smaller
than actual number of features. Experiments on eight publicly available
datasets show that our approach gives significant computational gains (both in
time and memory), over standard QPFS as well as other state of the art feature
selection methods, even while improving the overall accuracy.
|
Object detection via a multi-region & semantic segmentation-aware CNN
model | We propose an object detection system that relies on a multi-region deep
convolutional neural network (CNN) that also encodes semantic
segmentation-aware features. The resulting CNN-based representation aims at
capturing a diverse set of discriminative appearance factors and exhibits
localization sensitivity that is essential for accurate object localization. We
exploit the above properties of our recognition module by integrating it on an
iterative localization mechanism that alternates between scoring a box proposal
and refining its location with a deep CNN regression model. Thanks to the
efficient use of our modules, we detect objects with very high localization
accuracy. On the detection challenges of PASCAL VOC2007 and PASCAL VOC2012 we
achieve mAP of 78.2% and 73.9% correspondingly, surpassing any other published
work by a significant margin.
|
Optimal Decision-Theoretic Classification Using Non-Decomposable
Performance Metrics | We provide a general theoretical analysis of expected out-of-sample utility,
also referred to as decision-theoretic classification, for non-decomposable
binary classification metrics such as F-measure and Jaccard coefficient. Our
key result is that the expected out-of-sample utility for many performance
metrics is provably optimized by a classifier which is equivalent to a signed
thresholding of the conditional probability of the positive class. Our analysis
bridges a gap in the literature on binary classification, revealed in light of
recent results for non-decomposable metrics in population utility maximization
style classification. Our results identify checkable properties of a
performance metric which are sufficient to guarantee a probability ranking
principle. We propose consistent estimators for optimal expected out-of-sample
classification. As a consequence of the probability ranking principle,
computational requirements can be reduced from exponential to cubic complexity
in the general case, and further reduced to quadratic complexity in special
cases. We provide empirical results on simulated and benchmark datasets
evaluating the performance of the proposed algorithms for decision-theoretic
classification and comparing them to baseline and state-of-the-art methods in
population utility maximization for non-decomposable metrics.
|
Language Models for Image Captioning: The Quirks and What Works | Two recent approaches have achieved state-of-the-art results in image
captioning. The first uses a pipelined process where a set of candidate words
is generated by a convolutional neural network (CNN) trained on images, and
then a maximum entropy (ME) language model is used to arrange these words into
a coherent sentence. The second uses the penultimate activation layer of the
CNN as input to a recurrent neural network (RNN) that then generates the
caption sequence. In this paper, we compare the merits of these different
language modeling approaches for the first time by using the same
state-of-the-art CNN as input. We examine issues in the different approaches,
including linguistic irregularities, caption repetition, and data set overlap.
By combining key aspects of the ME and RNN methods, we achieve a new record
performance over previously published results on the benchmark COCO dataset.
However, the gains we see in BLEU do not translate to human judgments.
|
DART: Dropouts meet Multiple Additive Regression Trees | Multiple Additive Regression Trees (MART), an ensemble model of boosted
regression trees, is known to deliver high prediction accuracy for diverse
tasks, and it is widely used in practice. However, it suffers an issue which we
call over-specialization, wherein trees added at later iterations tend to
impact the prediction of only a few instances, and make negligible contribution
towards the remaining instances. This negatively affects the performance of the
model on unseen data, and also makes the model over-sensitive to the
contributions of the few, initially added tress. We show that the commonly used
tool to address this issue, that of shrinkage, alleviates the problem only to a
certain extent and the fundamental issue of over-specialization still remains.
In this work, we explore a different approach to address the problem that of
employing dropouts, a tool that has been recently proposed in the context of
learning deep neural networks. We propose a novel way of employing dropouts in
MART, resulting in the DART algorithm. We evaluate DART on ranking, regression
and classification tasks, using large scale, publicly available datasets, and
show that DART outperforms MART in each of the tasks, with a significant
margin. We also show that DART overcomes the issue of over-specialization to a
considerable extent.
|
An Asymptotically Optimal Policy for Uniform Bandits of Unknown Support | Consider the problem of a controller sampling sequentially from a finite
number of $N \geq 2$ populations, specified by random variables $X^i_k$, $ i =
1,\ldots , N,$ and $k = 1, 2, \ldots$; where $X^i_k$ denotes the outcome from
population $i$ the $k^{th}$ time it is sampled. It is assumed that for each
fixed $i$, $\{ X^i_k \}_{k \geq 1}$ is a sequence of i.i.d. uniform random
variables over some interval $[a_i, b_i]$, with the support (i.e., $a_i, b_i$)
unknown to the controller. The objective is to have a policy $\pi$ for
deciding, based on available data, from which of the $N$ populations to sample
from at any time $n=1,2,\ldots$ so as to maximize the expected sum of outcomes
of $n$ samples or equivalently to minimize the regret due to lack on
information of the parameters $\{ a_i \}$ and $\{ b_i \}$. In this paper, we
present a simple inflated sample mean (ISM) type policy that is asymptotically
optimal in the sense of its regret achieving the asymptotic lower bound of
Burnetas and Katehakis (1996). Additionally, finite horizon regret bounds are
given.
|
Deep Learning for Medical Image Segmentation | This report provides an overview of the current state of the art deep
learning architectures and optimisation techniques, and uses the ADNI
hippocampus MRI dataset as an example to compare the effectiveness and
efficiency of different convolutional architectures on the task of patch-based
3-dimensional hippocampal segmentation, which is important in the diagnosis of
Alzheimer's Disease. We found that a slightly unconventional "stacked 2D"
approach provides much better classification performance than simple 2D patches
without requiring significantly more computational power. We also examined the
popular "tri-planar" approach used in some recently published studies, and
found that it provides much better results than the 2D approaches, but also
with a moderate increase in computational power requirement. Finally, we
evaluated a full 3D convolutional architecture, and found that it provides
marginally better results than the tri-planar approach, but at the cost of a
very significant increase in computational power requirement.
|
Exploring Models and Data for Image Question Answering | This work aims to address the problem of image-based question-answering (QA)
with new models and datasets. In our work, we propose to use neural networks
and visual semantic embeddings, without intermediate stages such as object
detection and image segmentation, to predict answers to simple questions about
images. Our model performs 1.8 times better than the only published results on
an existing image QA dataset. We also present a question generation algorithm
that converts image descriptions, which are widely available, into QA form. We
used this algorithm to produce an order-of-magnitude larger dataset, with more
evenly distributed answers. A suite of baseline results on this new dataset are
also presented.
|
Human Social Interaction Modeling Using Temporal Deep Networks | We present a novel approach to computational modeling of social interactions
based on modeling of essential social interaction predicates (ESIPs) such as
joint attention and entrainment. Based on sound social psychological theory and
methodology, we collect a new "Tower Game" dataset consisting of audio-visual
capture of dyadic interactions labeled with the ESIPs. We expect this dataset
to provide a new avenue for research in computational social interaction
modeling. We propose a novel joint Discriminative Conditional Restricted
Boltzmann Machine (DCRBM) model that combines a discriminative component with
the generative power of CRBMs. Such a combination enables us to uncover
actionable constituents of the ESIPs in two steps. First, we train the DCRBM
model on the labeled data and get accurate (76\%-49\% across various ESIPs)
detection of the predicates. Second, we exploit the generative capability of
DCRBMs to activate the trained model so as to generate the lower-level data
corresponding to the specific ESIP that closely matches the actual training
data (with mean square error 0.01-0.1 for generating 100 frames). We are thus
able to decompose the ESIPs into their constituent actionable behaviors. Such a
purely computational determination of how to establish an ESIP such as
engagement is unprecedented.
|
Equitability, interval estimation, and statistical power | For analysis of a high-dimensional dataset, a common approach is to test a
null hypothesis of statistical independence on all variable pairs using a
non-parametric measure of dependence. However, because this approach attempts
to identify any non-trivial relationship no matter how weak, it often
identifies too many relationships to be useful. What is needed is a way of
identifying a smaller set of relationships that merit detailed further
analysis.
Here we formally present and characterize equitability, a property of
measures of dependence that aims to overcome this challenge. Notionally, an
equitable statistic is a statistic that, given some measure of noise, assigns
similar scores to equally noisy relationships of different types [Reshef et al.
2011]. We begin by formalizing this idea via a new object called the
interpretable interval, which functions as an interval estimate of the amount
of noise in a relationship of unknown type. We define an equitable statistic as
one with small interpretable intervals.
We then draw on the equivalence of interval estimation and hypothesis testing
to show that under moderate assumptions an equitable statistic is one that
yields well powered tests for distinguishing not only between trivial and
non-trivial relationships of all kinds but also between non-trivial
relationships of different strengths. This means that equitability allows us to
specify a threshold relationship strength $x_0$ and to search for relationships
of all kinds with strength greater than $x_0$. Thus, equitability can be
thought of as a strengthening of power against independence that enables
fruitful analysis of data sets with a small number of strong, interesting
relationships and a large number of weaker ones. We conclude with a
demonstration of how our two equivalent characterizations of equitability can
be used to evaluate the equitability of a statistic in practice.
|
Measuring dependence powerfully and equitably | Given a high-dimensional data set we often wish to find the strongest
relationships within it. A common strategy is to evaluate a measure of
dependence on every variable pair and retain the highest-scoring pairs for
follow-up. This strategy works well if the statistic used is equitable [Reshef
et al. 2015a], i.e., if, for some measure of noise, it assigns similar scores
to equally noisy relationships regardless of relationship type (e.g., linear,
exponential, periodic).
In this paper, we introduce and characterize a population measure of
dependence called MIC*. We show three ways that MIC* can be viewed: as the
population value of MIC, a highly equitable statistic from [Reshef et al.
2011], as a canonical "smoothing" of mutual information, and as the supremum of
an infinite sequence defined in terms of optimal one-dimensional partitions of
the marginals of the joint distribution. Based on this theory, we introduce an
efficient approach for computing MIC* from the density of a pair of random
variables, and we define a new consistent estimator MICe for MIC* that is
efficiently computable. In contrast, there is no known polynomial-time
algorithm for computing the original equitable statistic MIC. We show through
simulations that MICe has better bias-variance properties than MIC. We then
introduce and prove the consistency of a second statistic, TICe, that is a
trivial side-product of the computation of MICe and whose goal is powerful
independence testing rather than equitability.
We show in simulations that MICe and TICe have good equitability and power
against independence respectively. The analyses here complement a more in-depth
empirical evaluation of several leading measures of dependence [Reshef et al.
2015b] that shows state-of-the-art performance for MICe and TICe.
|
An Empirical Study of Leading Measures of Dependence | In exploratory data analysis, we are often interested in identifying
promising pairwise associations for further analysis while filtering out
weaker, less interesting ones. This can be accomplished by computing a measure
of dependence on all variable pairs and examining the highest-scoring pairs,
provided the measure of dependence used assigns similar scores to equally noisy
relationships of different types. This property, called equitability, is
formalized in Reshef et al. [2015b]. In addition to equitability, measures of
dependence can also be assessed by the power of their corresponding
independence tests as well as their runtime.
Here we present extensive empirical evaluation of the equitability, power
against independence, and runtime of several leading measures of dependence.
These include two statistics introduced in Reshef et al. [2015a]: MICe, which
has equitability as its primary goal, and TICe, which has power against
independence as its goal. Regarding equitability, our analysis finds that MICe
is the most equitable method on functional relationships in most of the
settings we considered, although mutual information estimation proves the most
equitable at large sample sizes in some specific settings. Regarding power
against independence, we find that TICe, along with Heller and Gorfine's S^DDP,
is the state of the art on the relationships we tested. Our analyses also show
a trade-off between power against independence and equitability consistent with
the theory in Reshef et al. [2015b]. In terms of runtime, MICe and TICe are
significantly faster than many other measures of dependence tested, and
computing either one makes computing the other trivial. This suggests that a
fast and useful strategy for achieving a combination of power against
independence and equitability may be to filter relationships by TICe and then
to examine the MICe of only the significant ones.
|
Newton Sketch: A Linear-time Optimization Algorithm with
Linear-Quadratic Convergence | We propose a randomized second-order method for optimization known as the
Newton Sketch: it is based on performing an approximate Newton step using a
randomly projected or sub-sampled Hessian. For self-concordant functions, we
prove that the algorithm has super-linear convergence with exponentially high
probability, with convergence and complexity guarantees that are independent of
condition numbers and related problem-dependent quantities. Given a suitable
initialization, similar guarantees also hold for strongly convex and smooth
objectives without self-concordance. When implemented using randomized
projections based on a sub-sampled Hadamard basis, the algorithm typically has
substantially lower complexity than Newton's method. We also describe
extensions of our methods to programs involving convex constraints that are
equipped with self-concordant barriers. We discuss and illustrate applications
to linear programs, quadratic programs with convex constraints, logistic
regression and other generalized linear models, as well as semidefinite
programs.
|
Probabilistic Cascading for Large Scale Hierarchical Classification | Hierarchies are frequently used for the organization of objects. Given a
hierarchy of classes, two main approaches are used, to automatically classify
new instances: flat classification and cascade classification. Flat
classification ignores the hierarchy, while cascade classification greedily
traverses the hierarchy from the root to the predicted leaf. In this paper we
propose a new approach, which extends cascade classification to predict the
right leaf by estimating the probability of each root-to-leaf path. We provide
experimental results which indicate that, using the same classification
algorithm, one can achieve better results with our approach, compared to the
traditional flat and cascade classifications.
|
Should we really use post-hoc tests based on mean-ranks? | The statistical comparison of multiple algorithms over multiple data sets is
fundamental in machine learning. This is typically carried out by the Friedman
test. When the Friedman test rejects the null hypothesis, multiple comparisons
are carried out to establish which are the significant differences among
algorithms. The multiple comparisons are usually performed using the mean-ranks
test. The aim of this technical note is to discuss the inconsistencies of the
mean-ranks post-hoc test with the goal of discouraging its use in machine
learning as well as in medicine, psychology, etc.. We show that the outcome of
the mean-ranks test depends on the pool of algorithms originally included in
the experiment. In other words, the outcome of the comparison between
algorithms A and B depends also on the performance of the other algorithms
included in the original experiment. This can lead to paradoxical situations.
For instance the difference between A and B could be declared significant if
the pool comprises algorithms C, D, E and not significant if the pool comprises
algorithms F, G, H. To overcome these issues, we suggest instead to perform the
multiple comparison using a test whose outcome only depends on the two
algorithms being compared, such as the sign-test or the Wilcoxon signed-rank
test.
|
Estimation with Norm Regularization | Analysis of non-asymptotic estimation error and structured statistical
recovery based on norm regularized regression, such as Lasso, needs to consider
four aspects: the norm, the loss function, the design matrix, and the noise
model. This paper presents generalizations of such estimation error analysis on
all four aspects compared to the existing literature. We characterize the
restricted error set where the estimation error vector lies, establish
relations between error sets for the constrained and regularized problems, and
present an estimation error bound applicable to any norm. Precise
characterizations of the bound is presented for isotropic as well as
anisotropic subGaussian design matrices, subGaussian noise models, and convex
loss functions, including least squares and generalized linear models. Generic
chaining and associated results play an important role in the analysis. A key
result from the analysis is that the sample complexity of all such estimators
depends on the Gaussian width of a spherical cap corresponding to the
restricted error set. Further, once the number of samples $n$ crosses the
required sample complexity, the estimation error decreases as
$\frac{c}{\sqrt{n}}$, where $c$ depends on the Gaussian width of the unit norm
ball.
|
Simultaneous Clustering and Model Selection for Multinomial
Distribution: A Comparative Study | In this paper, we study different discrete data clustering methods, which use
the Model-Based Clustering (MBC) framework with the Multinomial distribution.
Our study comprises several relevant issues, such as initialization, model
estimation and model selection. Additionally, we propose a novel MBC method by
efficiently combining the partitional and hierarchical clustering techniques.
We conduct experiments on both synthetic and real data and evaluate the methods
using accuracy, stability and computation time. Our study identifies
appropriate strategies to be used for discrete data analysis with the MBC
methods. Moreover, our proposed method is very competitive w.r.t. clustering
accuracy and better w.r.t. stability and computation time.
|
Bayesian Sparse Tucker Models for Dimension Reduction and Tensor
Completion | Tucker decomposition is the cornerstone of modern machine learning on
tensorial data analysis, which have attracted considerable attention for
multiway feature extraction, compressive sensing, and tensor completion. The
most challenging problem is related to determination of model complexity (i.e.,
multilinear rank), especially when noise and missing data are present. In
addition, existing methods cannot take into account uncertainty information of
latent factors, resulting in low generalization performance. To address these
issues, we present a class of probabilistic generative Tucker models for tensor
decomposition and completion with structural sparsity over multilinear latent
space. To exploit structural sparse modeling, we introduce two group sparsity
inducing priors by hierarchial representation of Laplace and Student-t
distributions, which facilitates fully posterior inference. For model learning,
we derived variational Bayesian inferences over all model (hyper)parameters,
and developed efficient and scalable algorithms based on multilinear
operations. Our methods can automatically adapt model complexity and infer an
optimal multilinear rank by the principle of maximum lower bound of model
evidence. Experimental results and comparisons on synthetic, chemometrics and
neuroimaging data demonstrate remarkable performance of our models for
recovering ground-truth of multilinear rank and missing entries.
|
Bounded-Distortion Metric Learning | Metric learning aims to embed one metric space into another to benefit tasks
like classification and clustering. Although a greatly distorted metric space
has a high degree of freedom to fit training data, it is prone to overfitting
and numerical inaccuracy. This paper presents {\it bounded-distortion metric
learning} (BDML), a new metric learning framework which amounts to finding an
optimal Mahalanobis metric space with a bounded-distortion constraint. An
efficient solver based on the multiplicative weights update method is proposed.
Moreover, we generalize BDML to pseudo-metric learning and devise the
semidefinite relaxation and a randomized algorithm to approximately solve it.
We further provide theoretical analysis to show that distortion is a key
ingredient for stability and generalization ability of our BDML algorithm.
Extensive experiments on several benchmark datasets yield promising results.
|
Towards stability and optimality in stochastic gradient descent | Iterative procedures for parameter estimation based on stochastic gradient
descent allow the estimation to scale to massive data sets. However, in both
theory and practice, they suffer from numerical instability. Moreover, they are
statistically inefficient as estimators of the true parameter value. To address
these two issues, we propose a new iterative procedure termed averaged implicit
SGD (AI-SGD). For statistical efficiency, AI-SGD employs averaging of the
iterates, which achieves the optimal Cram\'{e}r-Rao bound under strong
convexity, i.e., it is an optimal unbiased estimator of the true parameter
value. For numerical stability, AI-SGD employs an implicit update at each
iteration, which is related to proximal operators in optimization. In practice,
AI-SGD achieves competitive performance with other state-of-the-art procedures.
Furthermore, it is more stable than averaging procedures that do not employ
proximal updates, and is simple to implement as it requires fewer tunable
hyperparameters than procedures that do employ proximal updates.
|
Improved Relation Extraction with Feature-Rich Compositional Embedding
Models | Compositional embedding models build a representation (or embedding) for a
linguistic structure based on its component word embeddings. We propose a
Feature-rich Compositional Embedding Model (FCM) for relation extraction that
is expressive, generalizes to new domains, and is easy-to-implement. The key
idea is to combine both (unlexicalized) hand-crafted features with learned word
embeddings. The model is able to directly tackle the difficulties met by
traditional compositional embeddings models, such as handling arbitrary types
of sentence annotations and utilizing global information for composition. We
test the proposed model on two relation extraction tasks, and demonstrate that
our model outperforms both previous compositional models and traditional
feature rich models on the ACE 2005 relation extraction task, and the SemEval
2010 relation classification task. The combination of our model and a
log-linear classifier with hand-crafted features gives state-of-the-art
results.
|
Spike and Slab Gaussian Process Latent Variable Models | The Gaussian process latent variable model (GP-LVM) is a popular approach to
non-linear probabilistic dimensionality reduction. One design choice for the
model is the number of latent variables. We present a spike and slab prior for
the GP-LVM and propose an efficient variational inference procedure that gives
a lower bound of the log marginal likelihood. The new model provides a more
principled approach for selecting latent dimensions than the standard way of
thresholding the length-scale parameters. The effectiveness of our approach is
demonstrated through experiments on real and simulated data. Further, we extend
multi-view Gaussian processes that rely on sharing latent dimensions (known as
manifold relevance determination) with spike and slab priors. This allows a
more principled approach for selecting a subset of the latent space for each
view of data. The extended model outperforms the previous state-of-the-art when
applied to a cross-modal multimedia retrieval task.
|
Soft-Deep Boltzmann Machines | We present a layered Boltzmann machine (BM) that can better exploit the
advantages of a distributed representation. It is widely believed that deep BMs
(DBMs) have far greater representational power than its shallow counterpart,
restricted Boltzmann machines (RBMs). However, this expectation on the
supremacy of DBMs over RBMs has not ever been validated in a theoretical
fashion. In this paper, we provide both theoretical and empirical evidences
that the representational power of DBMs can be actually rather limited in
taking advantages of distributed representations. We propose an approximate
measure for the representational power of a BM regarding to the efficiency of a
distributed representation. With this measure, we show a surprising fact that
DBMs can make inefficient use of distributed representations. Based on these
observations, we propose an alternative BM architecture, which we dub soft-deep
BMs (sDBMs). We show that sDBMs can more efficiently exploit the distributed
representations in terms of the measure. Experiments demonstrate that sDBMs
outperform several state-of-the-art models, including DBMs, in generative tasks
on binarized MNIST and Caltech-101 silhouettes.
|
Improving neural networks with bunches of neurons modeled by Kumaraswamy
units: Preliminary study | Deep neural networks have recently achieved state-of-the-art results in many
machine learning problems, e.g., speech recognition or object recognition.
Hitherto, work on rectified linear units (ReLU) provides empirical and
theoretical evidence on performance increase of neural networks comparing to
typically used sigmoid activation function. In this paper, we investigate a new
manner of improving neural networks by introducing a bunch of copies of the
same neuron modeled by the generalized Kumaraswamy distribution. As a result,
we propose novel non-linear activation function which we refer to as
Kumaraswamy unit which is closely related to ReLU. In the experimental study
with MNIST image corpora we evaluate the Kumaraswamy unit applied to
single-layer (shallow) neural network and report a significant drop in test
classification error and test cross-entropy in comparison to sigmoid unit, ReLU
and Noisy ReLU.
|
Sample complexity of learning Mahalanobis distance metrics | Metric learning seeks a transformation of the feature space that enhances
prediction quality for the given task at hand. In this work we provide
PAC-style sample complexity rates for supervised metric learning. We give
matching lower- and upper-bounds showing that the sample complexity scales with
the representation dimension when no assumptions are made about the underlying
data distribution. However, by leveraging the structure of the data
distribution, we show that one can achieve rates that are fine-tuned to a
specific notion of intrinsic complexity for a given dataset. Our analysis
reveals that augmenting the metric learning optimization criterion with a
simple norm-based regularization can help adapt to a dataset's intrinsic
complexity, yielding better generalization. Experiments on benchmark datasets
validate our analysis and show that regularizing the metric can help discern
the signal even when the data contains high amounts of noise.
|
Asymptotic Behavior of Minimal-Exploration Allocation Policies: Almost
Sure, Arbitrarily Slow Growing Regret | The purpose of this paper is to provide further understanding into the
structure of the sequential allocation ("stochastic multi-armed bandit", or
MAB) problem by establishing probability one finite horizon bounds and
convergence rates for the sample (or "pseudo") regret associated with two
simple classes of allocation policies $\pi$.
For any slowly increasing function $g$, subject to mild regularity
constraints, we construct two policies (the $g$-Forcing, and the $g$-Inflated
Sample Mean) that achieve a measure of regret of order $ O(g(n))$ almost surely
as $n \to \infty$, bound from above and below. Additionally, almost sure upper
and lower bounds on the remainder term are established. In the constructions
herein, the function $g$ effectively controls the "exploration" of the
classical "exploration/exploitation" tradeoff.
|
The Boundary Forest Algorithm for Online Supervised and Unsupervised
Learning | We describe a new instance-based learning algorithm called the Boundary
Forest (BF) algorithm, that can be used for supervised and unsupervised
learning. The algorithm builds a forest of trees whose nodes store previously
seen examples. It can be shown data points one at a time and updates itself
incrementally, hence it is naturally online. Few instance-based algorithms have
this property while being simultaneously fast, which the BF is. This is crucial
for applications where one needs to respond to input data in real time. The
number of children of each node is not set beforehand but obtained from the
training procedure, which makes the algorithm very flexible with regards to
what data manifolds it can learn. We test its generalization performance and
speed on a range of benchmark datasets and detail in which settings it
outperforms the state of the art. Empirically we find that training time scales
as O(DNlog(N)) and testing as O(Dlog(N)), where D is the dimensionality and N
the amount of data,
|
Incorporating Type II Error Probabilities from Independence Tests into
Score-Based Learning of Bayesian Network Structure | We give a new consistent scoring function for structure learning of Bayesian
networks. In contrast to traditional approaches to score-based structure
learning, such as BDeu or MDL, the complexity penalty that we propose is
data-dependent and is given by the probability that a conditional independence
test correctly shows that an edge cannot exist. What really distinguishes this
new scoring function from earlier work is that it has the property of becoming
computationally easier to maximize as the amount of data increases. We prove a
polynomial sample complexity result, showing that maximizing this score is
guaranteed to correctly learn a structure with no false edges and a
distribution close to the generating distribution, whenever there exists a
Bayesian network which is a perfect map for the data generating distribution.
Although the new score can be used with any search algorithm, in our related
UAI 2013 paper [BS13], we have given empirical results showing that it is
particularly effective when used together with a linear programming relaxation
approach to Bayesian network structure learning. The present paper contains all
details of the proofs of the finite-sample complexity results in [BS13] as well
as detailed explanation of the computation of the certain error probabilities
called beta-values, whose precomputation and tabulation is necessary for the
implementation of the algorithm in [BS13].
|
Permutational Rademacher Complexity: a New Complexity Measure for
Transductive Learning | Transductive learning considers situations when a learner observes $m$
labelled training points and $u$ unlabelled test points with the final goal of
giving correct answers for the test points. This paper introduces a new
complexity measure for transductive learning called Permutational Rademacher
Complexity (PRC) and studies its properties. A novel symmetrization inequality
is proved, which shows that PRC provides a tighter control over expected
suprema of empirical processes compared to what happens in the standard i.i.d.
setting. A number of comparison results are also provided, which show the
relation between PRC and other popular complexity measures used in statistical
learning theory, including Rademacher complexity and Transductive Rademacher
Complexity (TRC). We argue that PRC is a more suitable complexity measure for
transductive learning. Finally, these results are combined with a standard
concentration argument to provide novel data-dependent risk bounds for
transductive learning.
|
Detecting the large entries of a sparse covariance matrix in
sub-quadratic time | The covariance matrix of a $p$-dimensional random variable is a fundamental
quantity in data analysis. Given $n$ i.i.d. observations, it is typically
estimated by the sample covariance matrix, at a computational cost of
$O(np^{2})$ operations. When $n,p$ are large, this computation may be
prohibitively slow. Moreover, in several contemporary applications, the
population matrix is approximately sparse, and only its few large entries are
of interest. This raises the following question, at the focus of our work:
Assuming approximate sparsity of the covariance matrix, can its large entries
be detected much faster, say in sub-quadratic time, without explicitly
computing all its $p^{2}$ entries? In this paper, we present and theoretically
analyze two randomized algorithms that detect the large entries of an
approximately sparse sample covariance matrix using only $O(np\text{ poly log }
p)$ operations. Furthermore, assuming sparsity of the population matrix, we
derive sufficient conditions on the underlying random variable and on the
number of samples $n$, for the sample covariance matrix to satisfy our
approximate sparsity requirements. Finally, we illustrate the performance of
our algorithms via several simulations.
|
Removing systematic errors for exoplanet search via latent causes | We describe a method for removing the effect of confounders in order to
reconstruct a latent quantity of interest. The method, referred to as
half-sibling regression, is inspired by recent work in causal inference using
additive noise models. We provide a theoretical justification and illustrate
the potential of the method in a challenging astronomy application.
|
Hybrid data clustering approach using K-Means and Flower Pollination
Algorithm | Data clustering is a technique for clustering set of objects into known
number of groups. Several approaches are widely applied to data clustering so
that objects within the clusters are similar and objects in different clusters
are far away from each other. K-Means, is one of the familiar center based
clustering algorithms since implementation is very easy and fast convergence.
However, K-Means algorithm suffers from initialization, hence trapped in local
optima. Flower Pollination Algorithm (FPA) is the global optimization
technique, which avoids trapping in local optimum solution. In this paper, a
novel hybrid data clustering approach using Flower Pollination Algorithm and
K-Means (FPAKM) is proposed. The proposed algorithm results are compared with
K-Means and FPA on eight datasets. From the experimental results, FPAKM is
better than FPA and K-Means.
|
Mind the duality gap: safer rules for the Lasso | Screening rules allow to early discard irrelevant variables from the
optimization in Lasso problems, or its derivatives, making solvers faster. In
this paper, we propose new versions of the so-called $\textit{safe rules}$ for
the Lasso. Based on duality gap considerations, our new rules create safe test
regions whose diameters converge to zero, provided that one relies on a
converging solver. This property helps screening out more variables, for a
wider range of regularization parameter values. In addition to faster
convergence, we prove that we correctly identify the active sets (supports) of
the solutions in finite time. While our proposed strategy can cope with any
solver, its performance is demonstrated using a coordinate descent algorithm
particularly adapted to machine learning use cases. Significant computing time
reductions are obtained with respect to previous safe rules.
|
Neural Network with Unbounded Activation Functions is Universal
Approximator | This paper presents an investigation of the approximation property of neural
networks with unbounded activation functions, such as the rectified linear unit
(ReLU), which is the new de-facto standard of deep learning. The ReLU network
can be analyzed by the ridgelet transform with respect to Lizorkin
distributions. By showing three reconstruction formulas by using the Fourier
slice theorem, the Radon transform, and Parseval's relation, it is shown that a
neural network with unbounded activation functions still satisfies the
universal approximation property. As an additional consequence, the ridgelet
transform, or the backprojection filter in the Radon domain, is what the
network learns after backpropagation. Subject to a constructive admissibility
condition, the trained network can be obtained by simply discretizing the
ridgelet transform, without backpropagation. Numerical examples not only
support the consistency of the admissibility condition but also imply that some
non-admissible cases result in low-pass filtering.
|
A PCA-Based Convolutional Network | In this paper, we propose a novel unsupervised deep learning model, called
PCA-based Convolutional Network (PCN). The architecture of PCN is composed of
several feature extraction stages and a nonlinear output stage. Particularly,
each feature extraction stage includes two layers: a convolutional layer and a
feature pooling layer. In the convolutional layer, the filter banks are simply
learned by PCA. In the nonlinear output stage, binary hashing is applied. For
the higher convolutional layers, the filter banks are learned from the feature
maps that were obtained in the previous stage. To test PCN, we conducted
extensive experiments on some challenging tasks, including handwritten digits
recognition, face recognition and texture classification. The results show that
PCN performs competitive with or even better than state-of-the-art deep
learning models. More importantly, since there is no back propagation for
supervised finetuning, PCN is much more efficient than existing deep networks.
|
Training generative neural networks via Maximum Mean Discrepancy
optimization | We consider training a deep neural network to generate samples from an
unknown distribution given i.i.d. data. We frame learning as an optimization
minimizing a two-sample test statistic---informally speaking, a good generator
network produces samples that cause a two-sample test to fail to reject the
null hypothesis. As our two-sample test statistic, we use an unbiased estimate
of the maximum mean discrepancy, which is the centerpiece of the nonparametric
kernel two-sample test proposed by Gretton et al. (2012). We compare to the
adversarial nets framework introduced by Goodfellow et al. (2014), in which
learning is a two-player game between a generator network and an adversarial
discriminator network, both trained to outwit the other. From this perspective,
the MMD statistic plays the role of the discriminator. In addition to empirical
comparisons, we prove bounds on the generalization error incurred by optimizing
the empirical MMD.
|
$k$-center Clustering under Perturbation Resilience | The $k$-center problem is a canonical and long-studied facility location and
clustering problem with many applications in both its symmetric and asymmetric
forms. Both versions of the problem have tight approximation factors on worst
case instances. Therefore to improve on these ratios, one must go beyond the
worst case.
In this work, we take this approach and provide strong positive results both
for the asymmetric and symmetric $k$-center problems under a natural input
stability (promise) condition called $\alpha$-perturbation resilience [Bilu and
Linia 2012], which states that the optimal solution does not change under any
alpha-factor perturbation to the input distances. We provide algorithms that
give strong guarantees simultaneously for stable and non-stable instances: our
algorithms always inherit the worst-case guarantees of clustering approximation
algorithms, and output the optimal solution if the input is $2$-perturbation
resilient. Furthermore, we prove our result is tight by showing symmetric
$k$-center under $(2-\epsilon)$-perturbation resilience is hard unless $NP=RP$.
The impact of our results are multifaceted. This is the first tight result
for any problem under perturbation resilience. Furthermore, our results
illustrate a surprising relationship between symmetric and asymmetric
$k$-center instances under perturbation resilience. Unlike approximation ratio,
for which symmetric $k$-center is easily solved to a factor of 2 but asymmetric
$k$-center cannot be approximated to any constant factor, both symmetric and
asymmetric $k$-center can be solved optimally under resilience to
2-perturbations. Finally, our guarantees in the setting where only part of the
data satisfies perturbation resilience makes these algorithms more applicable
to real-life instances.
|
Using Ensemble Models in the Histological Examination of Tissue
Abnormalities | Classification models for the automatic detection of abnormalities on
histological samples do exists, with an active debate on the cost associated
with false negative diagnosis (underdiagnosis) and false positive diagnosis
(overdiagnosis). Current models tend to underdiagnose, failing to recognize a
potentially fatal disease.
The objective of this study is to investigate the possibility of
automatically identifying abnormalities in tissue samples through the use of an
ensemble model on data generated by histological examination and to minimize
the number of false negative cases.
|
Safe Screening for Multi-Task Feature Learning with Multiple Data
Matrices | Multi-task feature learning (MTFL) is a powerful technique in boosting the
predictive performance by learning multiple related
classification/regression/clustering tasks simultaneously. However, solving the
MTFL problem remains challenging when the feature dimension is extremely large.
In this paper, we propose a novel screening rule---that is based on the dual
projection onto convex sets (DPC)---to quickly identify the inactive
features---that have zero coefficients in the solution vectors across all
tasks. One of the appealing features of DPC is that: it is safe in the sense
that the detected inactive features are guaranteed to have zero coefficients in
the solution vectors across all tasks. Thus, by removing the inactive features
from the training phase, we may have substantial savings in the computational
cost and memory usage without sacrificing accuracy. To the best of our
knowledge, it is the first screening rule that is applicable to sparse models
with multiple data matrices. A key challenge in deriving DPC is to solve a
nonconvex problem. We show that we can solve for the global optimum efficiently
via a properly chosen parametrization of the constraint set. Moreover, DPC has
very low computational cost and can be integrated with any existing solvers. We
have evaluated the proposed DPC rule on both synthetic and real data sets. The
experiments indicate that DPC is very effective in identifying the inactive
features---especially for high dimensional data---which leads to a speedup up
to several orders of magnitude.
|
Optimal Low-Rank Tensor Recovery from Separable Measurements: Four
Contractions Suffice | Tensors play a central role in many modern machine learning and signal
processing applications. In such applications, the target tensor is usually of
low rank, i.e., can be expressed as a sum of a small number of rank one
tensors. This motivates us to consider the problem of low rank tensor recovery
from a class of linear measurements called separable measurements. As specific
examples, we focus on two distinct types of separable measurement mechanisms
(a) Random projections, where each measurement corresponds to an inner product
of the tensor with a suitable random tensor, and (b) the completion problem
where measurements constitute revelation of a random set of entries. We present
a computationally efficient algorithm, with rigorous and order-optimal sample
complexity results (upto logarithmic factors) for tensor recovery. Our method
is based on reduction to matrix completion sub-problems and adaptation of
Leurgans' method for tensor decomposition. We extend the methodology and sample
complexity results to higher order tensors, and experimentally validate our
theoretical results.
|
MCODE: Multivariate Conditional Outlier Detection | Outlier detection aims to identify unusual data instances that deviate from
expected patterns. The outlier detection is particularly challenging when
outliers are context dependent and when they are defined by unusual
combinations of multiple outcome variable values. In this paper, we develop and
study a new conditional outlier detection approach for multivariate outcome
spaces that works by (1) transforming the conditional detection to the outlier
detection problem in a new (unconditional) space and (2) defining outlier
scores by analyzing the data in the new space. Our approach relies on the
classifier chain decomposition of the multi-dimensional classification problem
that lets us transform the output space into a probability vector, one
probability for each dimension of the output space. Outlier scores applied to
these transformed vectors are then used to detect the outliers. Experiments on
multiple multi-dimensional classification problems with the different outlier
injection rates show that our methodology is robust and able to successfully
identify outliers when outliers are either sparse (manifested in one or very
few dimensions) or dense (affecting multiple dimensions).
|
Margins, Kernels and Non-linear Smoothed Perceptrons | We focus on the problem of finding a non-linear classification function that
lies in a Reproducing Kernel Hilbert Space (RKHS) both from the primal point of
view (finding a perfect separator when one exists) and the dual point of view
(giving a certificate of non-existence), with special focus on generalizations
of two classical schemes - the Perceptron (primal) and Von-Neumann (dual)
algorithms.
We cast our problem as one of maximizing the regularized normalized
hard-margin ($\rho$) in an RKHS and %use the Representer Theorem to rephrase it
in terms of a Mahalanobis dot-product/semi-norm associated with the kernel's
(normalized and signed) Gram matrix. We derive an accelerated smoothed
algorithm with a convergence rate of $\tfrac{\sqrt {\log n}}{\rho}$ given $n$
separable points, which is strikingly similar to the classical kernelized
Perceptron algorithm whose rate is $\tfrac1{\rho^2}$. When no such classifier
exists, we prove a version of Gordan's separation theorem for RKHSs, and give a
reinterpretation of negative margins. This allows us to give guarantees for a
primal-dual algorithm that halts in $\min\{\tfrac{\sqrt n}{|\rho|},
\tfrac{\sqrt n}{\epsilon}\}$ iterations with a perfect separator in the RKHS if
the primal is feasible or a dual $\epsilon$-certificate of near-infeasibility.
|
Consistent Algorithms for Multiclass Classification with a Reject Option | We consider the problem of $n$-class classification ($n\geq 2$), where the
classifier can choose to abstain from making predictions at a given cost, say,
a factor $\alpha$ of the cost of misclassification. Designing consistent
algorithms for such $n$-class classification problems with a `reject option' is
the main goal of this paper, thereby extending and generalizing previously
known results for $n=2$. We show that the Crammer-Singer surrogate and the one
vs all hinge loss, albeit with a different predictor than the standard argmax,
yield consistent algorithms for this problem when $\alpha=\frac{1}{2}$. More
interestingly, we design a new convex surrogate that is also consistent for
this problem when $\alpha=\frac{1}{2}$ and operates on a much lower dimensional
space ($\log(n)$ as opposed to $n$). We also generalize all three surrogates to
be consistent for any $\alpha\in[0, \frac{1}{2}]$.
|
Algorithmic Connections Between Active Learning and Stochastic Convex
Optimization | Interesting theoretical associations have been established by recent papers
between the fields of active learning and stochastic convex optimization due to
the common role of feedback in sequential querying mechanisms. In this paper,
we continue this thread in two parts by exploiting these relations for the
first time to yield novel algorithms in both fields, further motivating the
study of their intersection. First, inspired by a recent optimization algorithm
that was adaptive to unknown uniform convexity parameters, we present a new
active learning algorithm for one-dimensional thresholds that can yield minimax
rates by adapting to unknown noise parameters. Next, we show that one can
perform $d$-dimensional stochastic minimization of smooth uniformly convex
functions when only granted oracle access to noisy gradient signs along any
coordinate instead of real-valued gradients, by using a simple randomized
coordinate descent procedure where each line search can be solved by
$1$-dimensional active learning, provably achieving the same error convergence
rate as having the entire real-valued gradient. Combining these two parts
yields an algorithm that solves stochastic convex optimization of uniformly
convex and smooth functions using only noisy gradient signs by repeatedly
performing active learning, achieves optimal rates and is adaptive to all
unknown convexity and smoothness parameters.
|
An Analysis of Active Learning With Uniform Feature Noise | In active learning, the user sequentially chooses values for feature $X$ and
an oracle returns the corresponding label $Y$. In this paper, we consider the
effect of feature noise in active learning, which could arise either because
$X$ itself is being measured, or it is corrupted in transmission to the oracle,
or the oracle returns the label of a noisy version of the query point. In
statistics, feature noise is known as "errors in variables" and has been
studied extensively in non-active settings. However, the effect of feature
noise in active learning has not been studied before. We consider the
well-known Berkson errors-in-variables model with additive uniform noise of
width $\sigma$.
Our simple but revealing setting is that of one-dimensional binary
classification setting where the goal is to learn a threshold (point where the
probability of a $+$ label crosses half). We deal with regression functions
that are antisymmetric in a region of size $\sigma$ around the threshold and
also satisfy Tsybakov's margin condition around the threshold. We prove minimax
lower and upper bounds which demonstrate that when $\sigma$ is smaller than the
minimiax active/passive noiseless error derived in \cite{CN07}, then noise has
no effect on the rates and one achieves the same noiseless rates. For larger
$\sigma$, the \textit{unflattening} of the regression function on convolution
with uniform noise, along with its local antisymmetry around the threshold,
together yield a behaviour where noise \textit{appears} to be beneficial. Our
key result is that active learning can buy significant improvement over a
passive strategy even in the presence of feature noise.
|
A New Perspective on Boosting in Linear Regression via Subgradient
Optimization and Relatives | In this paper we analyze boosting algorithms in linear regression from a new
perspective: that of modern first-order methods in convex optimization. We show
that classic boosting algorithms in linear regression, namely the incremental
forward stagewise algorithm (FS$_\varepsilon$) and least squares boosting
(LS-Boost($\varepsilon$)), can be viewed as subgradient descent to minimize the
loss function defined as the maximum absolute correlation between the features
and residuals. We also propose a modification of FS$_\varepsilon$ that yields
an algorithm for the Lasso, and that may be easily extended to an algorithm
that computes the Lasso path for different values of the regularization
parameter. Furthermore, we show that these new algorithms for the Lasso may
also be interpreted as the same master algorithm (subgradient descent), applied
to a regularized version of the maximum absolute correlation loss function. We
derive novel, comprehensive computational guarantees for several boosting
algorithms in linear regression (including LS-Boost($\varepsilon$) and
FS$_\varepsilon$) by using techniques of modern first-order methods in convex
optimization. Our computational guarantees inform us about the statistical
properties of boosting algorithms. In particular they provide, for the first
time, a precise theoretical description of the amount of data-fidelity and
regularization imparted by running a boosting algorithm with a prespecified
learning rate for a fixed but arbitrary number of iterations, for any dataset.
|
Global Convergence of Unmodified 3-Block ADMM for a Class of Convex
Minimization Problems | The alternating direction method of multipliers (ADMM) has been successfully
applied to solve structured convex optimization problems due to its superior
practical performance. The convergence properties of the 2-block ADMM have been
studied extensively in the literature. Specifically, it has been proven that
the 2-block ADMM globally converges for any penalty parameter $\gamma>0$. In
this sense, the 2-block ADMM allows the parameter to be free, i.e., there is no
need to restrict the value for the parameter when implementing this algorithm
in order to ensure convergence. However, for the 3-block ADMM, Chen \etal
\cite{Chen-admm-failure-2013} recently constructed a counter-example showing
that it can diverge if no further condition is imposed. The existing results on
studying further sufficient conditions on guaranteeing the convergence of the
3-block ADMM usually require $\gamma$ to be smaller than a certain bound, which
is usually either difficult to compute or too small to make it a practical
algorithm. In this paper, we show that the 3-block ADMM still globally
converges with any penalty parameter $\gamma>0$ if the third function $f_3$ in
the objective is smooth and strongly convex, and its condition number is in
$[1,1.0798)$, besides some other mild conditions. This requirement covers an
important class of problems to be called regularized least squares
decomposition (RLSD) in this paper.
|
Provably Correct Algorithms for Matrix Column Subset Selection with
Selectively Sampled Data | We consider the problem of matrix column subset selection, which selects a
subset of columns from an input matrix such that the input can be well
approximated by the span of the selected columns. Column subset selection has
been applied to numerous real-world data applications such as population
genetics summarization, electronic circuits testing and recommendation systems.
In many applications the complete data matrix is unavailable and one needs to
select representative columns by inspecting only a small portion of the input
matrix. In this paper we propose the first provably correct column subset
selection algorithms for partially observed data matrices. Our proposed
algorithms exhibit different merits and limitations in terms of statistical
accuracy, computational efficiency, sample complexity and sampling schemes,
which provides a nice exploration of the tradeoff between these desired
properties for column subset selection. The proposed methods employ the idea of
feedback driven sampling and are inspired by several sampling schemes
previously introduced for low-rank matrix approximation tasks (Drineas et al.,
2008; Frieze et al., 2004; Deshpande and Vempala, 2006; Krishnamurthy and
Singh, 2014). Our analysis shows that, under the assumption that the input data
matrix has incoherent rows but possibly coherent columns, all algorithms
provably converge to the best low-rank approximation of the original data as
number of selected columns increases. Furthermore, two of the proposed
algorithms enjoy a relative error bound, which is preferred for column subset
selection and matrix approximation purposes. We also demonstrate through both
theoretical and empirical analysis the power of feedback driven sampling
compared to uniform random sampling on input matrices with highly correlated
columns.
|
Shrinkage degree in $L_2$-re-scale boosting for regression | Re-scale boosting (RBoosting) is a variant of boosting which can essentially
improve the generalization performance of boosting learning. The key feature of
RBoosting lies in introducing a shrinkage degree to re-scale the ensemble
estimate in each gradient-descent step. Thus, the shrinkage degree determines
the performance of RBoosting.
The aim of this paper is to develop a concrete analysis concerning how to
determine the shrinkage degree in $L_2$-RBoosting. We propose two feasible ways
to select the shrinkage degree. The first one is to parameterize the shrinkage
degree and the other one is to develope a data-driven approach of it. After
rigorously analyzing the importance of the shrinkage degree in $L_2$-RBoosting
learning, we compare the pros and cons of the proposed methods. We find that
although these approaches can reach the same learning rates, the structure of
the final estimate of the parameterized approach is better, which sometimes
yields a better generalization capability when the number of sample is finite.
With this, we recommend to parameterize the shrinkage degree of
$L_2$-RBoosting. To this end, we present an adaptive parameter-selection
strategy for shrinkage degree and verify its feasibility through both
theoretical analysis and numerical verification.
The obtained results enhance the understanding of RBoosting and further give
guidance on how to use $L_2$-RBoosting for regression tasks.
|
Hinge-Loss Markov Random Fields and Probabilistic Soft Logic | A fundamental challenge in developing high-impact machine learning
technologies is balancing the need to model rich, structured domains with the
ability to scale to big data. Many important problem areas are both richly
structured and large scale, from social and biological networks, to knowledge
graphs and the Web, to images, video, and natural language. In this paper, we
introduce two new formalisms for modeling structured data, and show that they
can both capture rich structure and scale to big data. The first, hinge-loss
Markov random fields (HL-MRFs), is a new kind of probabilistic graphical model
that generalizes different approaches to convex inference. We unite three
approaches from the randomized algorithms, probabilistic graphical models, and
fuzzy logic communities, showing that all three lead to the same inference
objective. We then define HL-MRFs by generalizing this unified objective. The
second new formalism, probabilistic soft logic (PSL), is a probabilistic
programming language that makes HL-MRFs easy to define using a syntax based on
first-order logic. We introduce an algorithm for inferring most-probable
variable assignments (MAP inference) that is much more scalable than
general-purpose convex optimization methods, because it uses message passing to
take advantage of sparse dependency structures. We then show how to learn the
parameters of HL-MRFs. The learned HL-MRFs are as accurate as analogous
discrete models, but much more scalable. Together, these algorithms enable
HL-MRFs and PSL to model rich, structured data at scales not previously
possible.
|
Simple regret for infinitely many armed bandits | We consider a stochastic bandit problem with infinitely many arms. In this
setting, the learner has no chance of trying all the arms even once and has to
dedicate its limited number of samples only to a certain number of arms. All
previous algorithms for this setting were designed for minimizing the
cumulative regret of the learner. In this paper, we propose an algorithm aiming
at minimizing the simple regret. As in the cumulative regret setting of
infinitely many armed bandits, the rate of the simple regret will depend on a
parameter $\beta$ characterizing the distribution of the near-optimal arms. We
prove that depending on $\beta$, our algorithm is minimax optimal either up to
a multiplicative constant or up to a $\log(n)$ factor. We also provide
extensions to several important cases: when $\beta$ is unknown, in a natural
setting where the near-optimal arms have a small variance, and in the case of
unknown time horizon.
|
Recurrent Neural Network Training with Dark Knowledge Transfer | Recurrent neural networks (RNNs), particularly long short-term memory (LSTM),
have gained much attention in automatic speech recognition (ASR). Although some
successful stories have been reported, training RNNs remains highly
challenging, especially with limited training data. Recent research found that
a well-trained model can be used as a teacher to train other child models, by
using the predictions generated by the teacher model as supervision. This
knowledge transfer learning has been employed to train simple neural nets with
a complex one, so that the final performance can reach a level that is
infeasible to obtain by regular training. In this paper, we employ the
knowledge transfer learning approach to train RNNs (precisely LSTM) using a
deep neural network (DNN) model as the teacher. This is different from most of
the existing research on knowledge transfer learning, since the teacher (DNN)
is assumed to be weaker than the child (RNN); however, our experiments on an
ASR task showed that it works fairly well: without applying any tricks on the
learning scheme, this approach can train RNNs successfully even with limited
training data.
|
Graph Partitioning via Parallel Submodular Approximation to Accelerate
Distributed Machine Learning | Distributed computing excels at processing large scale data, but the
communication cost for synchronizing the shared parameters may slow down the
overall performance. Fortunately, the interactions between parameter and data
in many problems are sparse, which admits efficient partition in order to
reduce the communication overhead.
In this paper, we formulate data placement as a graph partitioning problem.
We propose a distributed partitioning algorithm. We give both theoretical
guarantees and a highly efficient implementation. We also provide a highly
efficient implementation of the algorithm and demonstrate its promising results
on both text datasets and social networks. We show that the proposed algorithm
leads to 1.6x speedup of a state-of-the-start distributed machine learning
system by eliminating 90\% of the network communication.
|
Ensemble of Example-Dependent Cost-Sensitive Decision Trees | Several real-world classification problems are example-dependent
cost-sensitive in nature, where the costs due to misclassification vary between
examples and not only within classes. However, standard classification methods
do not take these costs into account, and assume a constant cost of
misclassification errors. In previous works, some methods that take into
account the financial costs into the training of different algorithms have been
proposed, with the example-dependent cost-sensitive decision tree algorithm
being the one that gives the highest savings. In this paper we propose a new
framework of ensembles of example-dependent cost-sensitive decision-trees. The
framework consists in creating different example-dependent cost-sensitive
decision trees on random subsamples of the training set, and then combining
them using three different combination approaches. Moreover, we propose two new
cost-sensitive combination approaches; cost-sensitive weighted voting and
cost-sensitive stacking, the latter being based on the cost-sensitive logistic
regression method. Finally, using five different databases, from four
real-world applications: credit card fraud detection, churn modeling, credit
scoring and direct marketing, we evaluate the proposed method against
state-of-the-art example-dependent cost-sensitive techniques, namely,
cost-proportionate sampling, Bayes minimum risk and cost-sensitive decision
trees. The results show that the proposed algorithms have better results for
all databases, in the sense of higher savings.
|
Compressed Nonnegative Matrix Factorization is Fast and Accurate | Nonnegative matrix factorization (NMF) has an established reputation as a
useful data analysis technique in numerous applications. However, its usage in
practical situations is undergoing challenges in recent years. The fundamental
factor to this is the increasingly growing size of the datasets available and
needed in the information sciences. To address this, in this work we propose to
use structured random compression, that is, random projections that exploit the
data structure, for two NMF variants: classical and separable. In separable NMF
(SNMF) the left factors are a subset of the columns of the input matrix. We
present suitable formulations for each problem, dealing with different
representative algorithms within each one. We show that the resulting
compressed techniques are faster than their uncompressed variants, vastly
reduce memory demands, and do not encompass any significant deterioration in
performance. The proposed structured random projections for SNMF allow to deal
with arbitrarily shaped large matrices, beyond the standard limit of
tall-and-skinny matrices, granting access to very efficient computations in
this general setting. We accompany the algorithmic presentation with
theoretical foundations and numerous and diverse examples, showing the
suitability of the proposed approaches.
|
Layered Adaptive Importance Sampling | Monte Carlo methods represent the "de facto" standard for approximating
complicated integrals involving multidimensional target distributions. In order
to generate random realizations from the target distribution, Monte Carlo
techniques use simpler proposal probability densities to draw candidate
samples. The performance of any such method is strictly related to the
specification of the proposal distribution, such that unfortunate choices
easily wreak havoc on the resulting estimators. In this work, we introduce a
layered (i.e., hierarchical) procedure to generate samples employed within a
Monte Carlo scheme. This approach ensures that an appropriate equivalent
proposal density is always obtained automatically (thus eliminating the risk of
a catastrophic performance), although at the expense of a moderate increase in
the complexity. Furthermore, we provide a general unified importance sampling
(IS) framework, where multiple proposal densities are employed and several IS
schemes are introduced by applying the so-called deterministic mixture
approach. Finally, given these schemes, we also propose a novel class of
adaptive importance samplers using a population of proposals, where the
adaptation is driven by independent parallel or interacting Markov Chain Monte
Carlo (MCMC) chains. The resulting algorithms efficiently combine the benefits
of both IS and MCMC methods.
|
DopeLearning: A Computational Approach to Rap Lyrics Generation | Writing rap lyrics requires both creativity to construct a meaningful,
interesting story and lyrical skills to produce complex rhyme patterns, which
form the cornerstone of good flow. We present a rap lyrics generation method
that captures both of these aspects. First, we develop a prediction model to
identify the next line of existing lyrics from a set of candidate next lines.
This model is based on two machine-learning techniques: the RankSVM algorithm
and a deep neural network model with a novel structure. Results show that the
prediction model can identify the true next line among 299 randomly selected
lines with an accuracy of 17%, i.e., over 50 times more likely than by random.
Second, we employ the prediction model to combine lines from existing songs,
producing lyrics with rhyme and a meaning. An evaluation of the produced lyrics
shows that in terms of quantitative rhyme density, the method outperforms the
best human rappers by 21%. The rap lyrics generator has been deployed as an
online tool called DeepBeat, and the performance of the tool has been assessed
by analyzing its usage logs. This analysis shows that machine-learned rankings
correlate with user preferences.
|
On the tightness of an SDP relaxation of k-means | Recently, Awasthi et al. introduced an SDP relaxation of the $k$-means
problem in $\mathbb R^m$. In this work, we consider a random model for the data
points in which $k$ balls of unit radius are deterministically distributed
throughout $\mathbb R^m$, and then in each ball, $n$ points are drawn according
to a common rotationally invariant probability distribution. For any fixed ball
configuration and probability distribution, we prove that the SDP relaxation of
the $k$-means problem exactly recovers these planted clusters with probability
$1-e^{-\Omega(n)}$ provided the distance between any two of the ball centers is
$>2+\epsilon$, where $\epsilon$ is an explicit function of the configuration of
the ball centers, and can be arbitrarily small when $m$ is large.
|
Multi-task additive models with shared transfer functions based on
dictionary learning | Additive models form a widely popular class of regression models which
represent the relation between covariates and response variables as the sum of
low-dimensional transfer functions. Besides flexibility and accuracy, a key
benefit of these models is their interpretability: the transfer functions
provide visual means for inspecting the models and identifying domain-specific
relations between inputs and outputs. However, in large-scale problems
involving the prediction of many related tasks, learning independently additive
models results in a loss of model interpretability, and can cause overfitting
when training data is scarce. We introduce a novel multi-task learning approach
which provides a corpus of accurate and interpretable additive models for a
large number of related forecasting tasks. Our key idea is to share transfer
functions across models in order to reduce the model complexity and ease the
exploration of the corpus. We establish a connection with sparse dictionary
learning and propose a new efficient fitting algorithm which alternates between
sparse coding and transfer function updates. The former step is solved via an
extension of Orthogonal Matching Pursuit, whose properties are analyzed using a
novel recovery condition which extends existing results in the literature. The
latter step is addressed using a traditional dictionary update rule.
Experiments on real-world data demonstrate that our approach compares favorably
to baseline methods while yielding an interpretable corpus of models, revealing
structure among the individual tasks and being more robust when training data
is scarce. Our framework therefore extends the well-known benefits of additive
models to common regression settings possibly involving thousands of tasks.
|
Risk and Regret of Hierarchical Bayesian Learners | Common statistical practice has shown that the full power of Bayesian methods
is not realized until hierarchical priors are used, as these allow for greater
"robustness" and the ability to "share statistical strength." Yet it is an
ongoing challenge to provide a learning-theoretically sound formalism of such
notions that: offers practical guidance concerning when and how best to utilize
hierarchical models; provides insights into what makes for a good hierarchical
prior; and, when the form of the prior has been chosen, can guide the choice of
hyperparameter settings. We present a set of analytical tools for understanding
hierarchical priors in both the online and batch learning settings. We provide
regret bounds under log-loss, which show how certain hierarchical models
compare, in retrospect, to the best single model in the model class. We also
show how to convert a Bayesian log-loss regret bound into a Bayesian risk bound
for any bounded loss, a result which may be of independent interest. Risk and
regret bounds for Student's $t$ and hierarchical Gaussian priors allow us to
formalize the concepts of "robustness" and "sharing statistical strength."
Priors for feature selection are investigated as well. Our results suggest that
the learning-theoretic benefits of using hierarchical priors can often come at
little cost on practical problems.
|
An Experimental Comparison of Hybrid Algorithms for Bayesian Network
Structure Learning | We present a novel hybrid algorithm for Bayesian network structure learning,
called Hybrid HPC (H2PC). It first reconstructs the skeleton of a Bayesian
network and then performs a Bayesian-scoring greedy hill-climbing search to
orient the edges. It is based on a subroutine called HPC, that combines ideas
from incremental and divide-and-conquer constraint-based methods to learn the
parents and children of a target variable. We conduct an experimental
comparison of H2PC against Max-Min Hill-Climbing (MMHC), which is currently the
most powerful state-of-the-art algorithm for Bayesian network structure
learning, on several benchmarks with various data sizes. Our extensive
experiments show that H2PC outperforms MMHC both in terms of goodness of fit to
new data and in terms of the quality of the network structure itself, which is
closer to the true dependence structure of the data. The source code (in R) of
H2PC as well as all data sets used for the empirical tests are publicly
available.
|
Modelling-based experiment retrieval: A case study with gene expression
clustering | Motivation: Public and private repositories of experimental data are growing
to sizes that require dedicated methods for finding relevant data. To improve
on the state of the art of keyword searches from annotations, methods for
content-based retrieval have been proposed. In the context of gene expression
experiments, most methods retrieve gene expression profiles, requiring each
experiment to be expressed as a single profile, typically of case vs. control.
A more general, recently suggested alternative is to retrieve experiments whose
models are good for modelling the query dataset. However, for very noisy and
high-dimensional query data, this retrieval criterion turns out to be very
noisy as well.
Results: We propose doing retrieval using a denoised model of the query
dataset, instead of the original noisy dataset itself. To this end, we
introduce a general probabilistic framework, where each experiment is modelled
separately and the retrieval is done by finding related models. For retrieval
of gene expression experiments, we use a probabilistic model called product
partition model, which induces a clustering of genes that show similar
expression patterns across a number of samples. The suggested metric for
retrieval using clusterings is the normalized information distance. Empirical
results finally suggest that inference for the full probabilistic model can be
approximated with good performance using computationally faster heuristic
clustering approaches (e.g. $k$-means). The method is highly scalable and
straightforward to apply to construct a general-purpose gene expression
experiment retrieval method.
Availability: The method can be implemented using standard clustering
algorithms and normalized information distance, available in many statistical
software packages.
|
Solving Random Quadratic Systems of Equations Is Nearly as Easy as
Solving Linear Systems | We consider the fundamental problem of solving quadratic systems of equations
in $n$ variables, where $y_i = |\langle \boldsymbol{a}_i, \boldsymbol{x}
\rangle|^2$, $i = 1, \ldots, m$ and $\boldsymbol{x} \in \mathbb{R}^n$ is
unknown. We propose a novel method, which starting with an initial guess
computed by means of a spectral method, proceeds by minimizing a nonconvex
functional as in the Wirtinger flow approach. There are several key
distinguishing features, most notably, a distinct objective functional and
novel update rules, which operate in an adaptive fashion and drop terms bearing
too much influence on the search direction. These careful selection rules
provide a tighter initial guess, better descent directions, and thus enhanced
practical performance. On the theoretical side, we prove that for certain
unstructured models of quadratic systems, our algorithms return the correct
solution in linear time, i.e. in time proportional to reading the data
$\{\boldsymbol{a}_i\}$ and $\{y_i\}$ as soon as the ratio $m/n$ between the
number of equations and unknowns exceeds a fixed numerical constant. We extend
the theory to deal with noisy systems in which we only have $y_i \approx
|\langle \boldsymbol{a}_i, \boldsymbol{x} \rangle|^2$ and prove that our
algorithms achieve a statistical accuracy, which is nearly un-improvable. We
complement our theoretical study with numerical examples showing that solving
random quadratic systems is both computationally and statistically not much
harder than solving linear systems of the same size---hence the title of this
paper. For instance, we demonstrate empirically that the computational cost of
our algorithm is about four times that of solving a least-squares problem of
the same size.
|
oASIS: Adaptive Column Sampling for Kernel Matrix Approximation | Kernel matrices (e.g. Gram or similarity matrices) are essential for many
state-of-the-art approaches to classification, clustering, and dimensionality
reduction. For large datasets, the cost of forming and factoring such kernel
matrices becomes intractable. To address this challenge, we introduce a new
adaptive sampling algorithm called Accelerated Sequential Incoherence Selection
(oASIS) that samples columns without explicitly computing the entire kernel
matrix. We provide conditions under which oASIS is guaranteed to exactly
recover the kernel matrix with an optimal number of columns selected. Numerical
experiments on both synthetic and real-world datasets demonstrate that oASIS
achieves performance comparable to state-of-the-art adaptive sampling methods
at a fraction of the computational cost. The low runtime complexity of oASIS
and its low memory footprint enable the solution of large problems that are
simply intractable using other adaptive methods.
|
Learning with a Drifting Target Concept | We study the problem of learning in the presence of a drifting target
concept. Specifically, we provide bounds on the error rate at a given time,
given a learner with access to a history of independent samples labeled
according to a target concept that can change on each round. One of our main
contributions is a refinement of the best previous results for polynomial-time
algorithms for the space of linear separators under a uniform distribution. We
also provide general results for an algorithm capable of adapting to a variable
rate of drift of the target concept. Some of the results also describe an
active learning variant of this setting, and provide bounds on the number of
queries for the labels of points in the sequence sufficient to obtain the
stated bounds on the error rates.
|
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.