title
stringlengths 7
246
| abstract
stringlengths 6
3.31k
|
|---|---|
A Generalized Markov-Chain Modelling Approach to $(1,\lambda)$-ES Linear
Optimization: Technical Report | Several recent publications investigated Markov-chain modelling of linear
optimization by a $(1,\lambda)$-ES, considering both unconstrained and linearly
constrained optimization, and both constant and varying step size. All of them
assume normality of the involved random steps, and while this is consistent
with a black-box scenario, information on the function to be optimized (e.g.
separability) may be exploited by the use of another distribution. The
objective of our contribution is to complement previous studies realized with
normal steps, and to give sufficient conditions on the distribution of the
random steps for the success of a constant step-size $(1,\lambda)$-ES on the
simple problem of a linear function with a linear constraint. The decomposition
of a multidimensional distribution into its marginals and the copula combining
them is applied to the new distributional assumptions, particular attention
being paid to distributions with Archimedean copulas.
|
An Entropy Search Portfolio for Bayesian Optimization | Bayesian optimization is a sample-efficient method for black-box global
optimization. How- ever, the performance of a Bayesian optimization method very
much depends on its exploration strategy, i.e. the choice of acquisition
function, and it is not clear a priori which choice will result in superior
performance. While portfolio methods provide an effective, principled way of
combining a collection of acquisition functions, they are often based on
measures of past performance which can be misleading. To address this issue, we
introduce the Entropy Search Portfolio (ESP): a novel approach to portfolio
construction which is motivated by information theoretic considerations. We
show that ESP outperforms existing portfolio methods on several real and
synthetic problems, including geostatistical datasets and simulated control
tasks. We not only show that ESP is able to offer performance as good as the
best, but unknown, acquisition function, but surprisingly it often gives better
performance. Finally, over a wide range of conditions we find that ESP is
robust to the inclusion of poor acquisition functions.
|
A Sober Look at Spectral Learning | Spectral learning recently generated lots of excitement in machine learning,
largely because it is the first known method to produce consistent estimates
(under suitable conditions) for several latent variable models. In contrast,
maximum likelihood estimates may get trapped in local optima due to the
non-convex nature of the likelihood function of latent variable models. In this
paper, we do an empirical evaluation of spectral learning (SL) and expectation
maximization (EM), which reveals an important gap between the theory and the
practice. First, SL often leads to negative probabilities. Second, EM often
yields better estimates than spectral learning and it does not seem to get
stuck in local optima. We discuss how the rank of the model parameters and the
amount of training data can yield negative probabilities. We also question the
common belief that maximum likelihood estimators are necessarily inconsistent.
|
Exact Decoding on Latent Variable Conditional Models is NP-Hard | Latent variable conditional models, including the latent conditional random
fields as a special case, are popular models for many natural language
processing and vision processing tasks. The computational complexity of the
exact decoding/inference in latent conditional random fields is unclear. In
this paper, we try to clarify the computational complexity of the exact
decoding. We analyze the complexity and demonstrate that it is an NP-hard
problem even on a sequential labeling setting. Furthermore, we propose the
latent-dynamic inference (LDI-Naive) method and its bounded version
(LDI-Bounded), which are able to perform exact-inference or
almost-exact-inference by using top-$n$ search and dynamic programming.
|
An Experimental Evaluation of Nearest Neighbour Time Series
Classification | Data mining research into time series classification (TSC) has focussed on
alternative distance measures for nearest neighbour classifiers. It is standard
practice to use 1-NN with Euclidean or dynamic time warping (DTW) distance as a
straw man for comparison. As part of a wider investigation into elastic
distance measures for TSC~\cite{lines14elastic}, we perform a series of
experiments to test whether this standard practice is valid.
Specifically, we compare 1-NN classifiers with Euclidean and DTW distance to
standard classifiers, examine whether the performance of 1-NN Euclidean
approaches that of 1-NN DTW as the number of cases increases, assess whether
there is any benefit of setting $k$ for $k$-NN through cross validation whether
it is worth setting the warping path for DTW through cross validation and
finally is it better to use a window or weighting for DTW. Based on experiments
on 77 problems, we conclude that 1-NN with Euclidean distance is fairly easy to
beat but 1-NN with DTW is not, if window size is set through cross validation.
|
Predictive Modelling of Bone Age through Classification and Regression
of Bone Shapes | Bone age assessment is a task performed daily in hospitals worldwide. This
involves a clinician estimating the age of a patient from a radiograph of the
non-dominant hand.
Our approach to automated bone age assessment is to modularise the algorithm
into the following three stages: segment and verify hand outline; segment and
verify bones; use the bone outlines to construct models of age. In this paper
we address the final question: given outlines of bones, can we learn how to
predict the bone age of the patient? We examine two alternative approaches.
Firstly, we attempt to train classifiers on individual bones to predict the
bone stage categories commonly used in bone ageing. Secondly, we construct
regression models to directly predict patient age.
We demonstrate that models built on summary features of the bone outline
perform better than those built using the one dimensional representation of the
outline, and also do at least as well as other automated systems. We show that
models constructed on just three bones are as accurate at predicting age as
expert human assessors using the standard technique. We also demonstrate the
utility of the model by quantifying the importance of ethnicity and sex on age
development. Our conclusion is that the feature based system of separating the
image processing from the age modelling is the best approach for automated bone
ageing, since it offers flexibility and transparency and produces accurate
estimates.
|
Improved Densification of One Permutation Hashing | The existing work on densification of one permutation hashing reduces the
query processing cost of the $(K,L)$-parameterized Locality Sensitive Hashing
(LSH) algorithm with minwise hashing, from $O(dKL)$ to merely $O(d + KL)$,
where $d$ is the number of nonzeros of the data vector, $K$ is the number of
hashes in each hash table, and $L$ is the number of hash tables. While that is
a substantial improvement, our analysis reveals that the existing densification
scheme is sub-optimal. In particular, there is no enough randomness in that
procedure, which affects its accuracy on very sparse datasets.
In this paper, we provide a new densification procedure which is provably
better than the existing scheme. This improvement is more significant for very
sparse datasets which are common over the web. The improved technique has the
same cost of $O(d + KL)$ for query processing, thereby making it strictly
preferable over the existing procedure. Experimental evaluations on public
datasets, in the task of hashing based near neighbor search, support our
theoretical findings.
|
Homotopy based algorithms for $\ell_0$-regularized least-squares | Sparse signal restoration is usually formulated as the minimization of a
quadratic cost function $\|y-Ax\|_2^2$, where A is a dictionary and x is an
unknown sparse vector. It is well-known that imposing an $\ell_0$ constraint
leads to an NP-hard minimization problem. The convex relaxation approach has
received considerable attention, where the $\ell_0$-norm is replaced by the
$\ell_1$-norm. Among the many efficient $\ell_1$ solvers, the homotopy
algorithm minimizes $\|y-Ax\|_2^2+\lambda\|x\|_1$ with respect to x for a
continuum of $\lambda$'s. It is inspired by the piecewise regularity of the
$\ell_1$-regularization path, also referred to as the homotopy path. In this
paper, we address the minimization problem $\|y-Ax\|_2^2+\lambda\|x\|_0$ for a
continuum of $\lambda$'s and propose two heuristic search algorithms for
$\ell_0$-homotopy. Continuation Single Best Replacement is a forward-backward
greedy strategy extending the Single Best Replacement algorithm, previously
proposed for $\ell_0$-minimization at a given $\lambda$. The adaptive search of
the $\lambda$-values is inspired by $\ell_1$-homotopy. $\ell_0$ Regularization
Path Descent is a more complex algorithm exploiting the structural properties
of the $\ell_0$-regularization path, which is piecewise constant with respect
to $\lambda$. Both algorithms are empirically evaluated for difficult inverse
problems involving ill-conditioned dictionaries. Finally, we show that they can
be easily coupled with usual methods of model order selection.
|
On the Application of Generic Summarization Algorithms to Music | Several generic summarization algorithms were developed in the past and
successfully applied in fields such as text and speech summarization. In this
paper, we review and apply these algorithms to music. To evaluate this
summarization's performance, we adopt an extrinsic approach: we compare a Fado
Genre Classifier's performance using truncated contiguous clips against the
summaries extracted with those algorithms on 2 different datasets. We show that
Maximal Marginal Relevance (MMR), LexRank and Latent Semantic Analysis (LSA)
all improve classification performance in both datasets used for testing.
|
Variational Gaussian Process State-Space Models | State-space models have been successfully used for more than fifty years in
different areas of science and engineering. We present a procedure for
efficient variational Bayesian learning of nonlinear state-space models based
on sparse Gaussian processes. The result of learning is a tractable posterior
over nonlinear dynamical systems. In comparison to conventional parametric
models, we offer the possibility to straightforwardly trade off model capacity
and computational cost whilst avoiding overfitting. Our main algorithm uses a
hybrid inference approach combining variational Bayes and sequential Monte
Carlo. We also present stochastic variational inference and online learning
approaches for fast learning with long time series.
|
Brain-like associative learning using a nanoscale non-volatile phase
change synaptic device array | Recent advances in neuroscience together with nanoscale electronic device
technology have resulted in huge interests in realizing brain-like computing
hardwares using emerging nanoscale memory devices as synaptic elements.
Although there has been experimental work that demonstrated the operation of
nanoscale synaptic element at the single device level, network level studies
have been limited to simulations. In this work, we demonstrate, using
experiments, array level associative learning using phase change synaptic
devices connected in a grid like configuration similar to the organization of
the biological brain. Implementing Hebbian learning with phase change memory
cells, the synaptic grid was able to store presented patterns and recall
missing patterns in an associative brain-like fashion. We found that the system
is robust to device variations, and large variations in cell resistance states
can be accommodated by increasing the number of training epochs. We illustrated
the tradeoff between variation tolerance of the network and the overall energy
consumption, and found that energy consumption is decreased significantly for
lower variation tolerance.
|
Inner Product Similarity Search using Compositional Codes | This paper addresses the nearest neighbor search problem under inner product
similarity and introduces a compact code-based approach. The idea is to
approximate a vector using the composition of several elements selected from a
source dictionary and to represent this vector by a short code composed of the
indices of the selected elements. The inner product between a query vector and
a database vector is efficiently estimated from the query vector and the short
code of the database vector. We show the superior performance of the proposed
group $M$-selection algorithm that selects $M$ elements from $M$ source
dictionaries for vector approximation in terms of search accuracy and
efficiency for compact codes of the same length via theoretical and empirical
analysis. Experimental results on large-scale datasets ($1M$ and $1B$ SIFT
features, $1M$ linear models and Netflix) demonstrate the superiority of the
proposed approach.
|
Inferring causal structure: a quantum advantage | The problem of using observed correlations to infer causal relations is
relevant to a wide variety of scientific disciplines. Yet given correlations
between just two classical variables, it is impossible to determine whether
they arose from a causal influence of one on the other or a common cause
influencing both, unless one can implement a randomized intervention. We here
consider the problem of causal inference for quantum variables. We introduce
causal tomography, which unifies and generalizes conventional quantum
tomography schemes to provide a complete solution to the causal inference
problem using a quantum analogue of a randomized trial. We furthermore show
that, in contrast to the classical case, observed quantum correlations alone
can sometimes provide a solution. We implement a quantum-optical experiment
that allows us to control the causal relation between two optical modes, and
two measurement schemes -- one with and one without randomization -- that
extract this relation from the observed correlations. Our results show that
entanglement and coherence, known to be central to quantum information
processing, also provide a quantum advantage for causal inference.
|
The Sample Complexity of Learning Linear Predictors with the Squared
Loss | In this short note, we provide a sample complexity lower bound for learning
linear predictors with respect to the squared loss. Our focus is on an agnostic
setting, where no assumptions are made on the data distribution. This contrasts
with standard results in the literature, which either make distributional
assumptions, refer to specific parameter settings, or use other performance
measures.
|
Fast Support Vector Machines Using Parallel Adaptive Shrinking on
Distributed Systems | Support Vector Machines (SVM), a popular machine learning technique, has been
applied to a wide range of domains such as science, finance, and social
networks for supervised learning. Whether it is identifying high-risk patients
by health-care professionals, or potential high-school students to enroll in
college by school districts, SVMs can play a major role for social good. This
paper undertakes the challenge of designing a scalable parallel SVM training
algorithm for large scale systems, which includes commodity multi-core
machines, tightly connected supercomputers and cloud computing systems.
Intuitive techniques for improving the time-space complexity including adaptive
elimination of samples for faster convergence and sparse format representation
are proposed. Under sample elimination, several heuristics for {\em earliest
possible} to {\em lazy} elimination of non-contributing samples are proposed.
In several cases, where an early sample elimination might result in a false
positive, low overhead mechanisms for reconstruction of key data structures are
proposed. The algorithm and heuristics are implemented and evaluated on various
publicly available datasets. Empirical evaluation shows up to 26x speed
improvement on some datasets against the sequential baseline, when evaluated on
multiple compute nodes, and an improvement in execution time up to 30-60\% is
readily observed on a number of other datasets against our parallel baseline.
|
Enhancing Pure-Pixel Identification Performance via Preconditioning | In this paper, we analyze different preconditionings designed to enhance
robustness of pure-pixel search algorithms, which are used for blind
hyperspectral unmixing and which are equivalent to near-separable nonnegative
matrix factorization algorithms. Our analysis focuses on the successive
projection algorithm (SPA), a simple, efficient and provably robust algorithm
in the pure-pixel algorithm class. Recently, a provably robust preconditioning
was proposed by Gillis and Vavasis (arXiv:1310.2273) which requires the
resolution of a semidefinite program (SDP) to find a data points-enclosing
minimum volume ellipsoid. Since solving the SDP in high precisions can be time
consuming, we generalize the robustness analysis to approximate solutions of
the SDP, that is, solutions whose objective function values are some
multiplicative factors away from the optimal value. It is shown that a high
accuracy solution is not crucial for robustness, which paves the way for faster
preconditionings (e.g., based on first-order optimization methods). This first
contribution also allows us to provide a robustness analysis for two other
preconditionings. The first one is pre-whitening, which can be interpreted as
an optimal solution of the same SDP with additional constraints. We analyze
robustness of pre-whitening which allows us to characterize situations in which
it performs competitively with the SDP-based preconditioning. The second one is
based on SPA itself and can be interpreted as an optimal solution of a
relaxation of the SDP. It is extremely fast while competing with the SDP-based
preconditioning on several synthetic data sets.
|
Generalized Dantzig Selector: Application to the k-support norm | We propose a Generalized Dantzig Selector (GDS) for linear models, in which
any norm encoding the parameter structure can be leveraged for estimation. We
investigate both computational and statistical aspects of the GDS. Based on
conjugate proximal operator, a flexible inexact ADMM framework is designed for
solving GDS, and non-asymptotic high-probability bounds are established on the
estimation error, which rely on Gaussian width of unit norm ball and suitable
set encompassing estimation error. Further, we consider a non-trivial example
of the GDS using $k$-support norm. We derive an efficient method to compute the
proximal operator for $k$-support norm since existing methods are inapplicable
in this setting. For statistical analysis, we provide upper bounds for the
Gaussian widths needed in the GDS analysis, yielding the first statistical
recovery guarantee for estimation with the $k$-support norm. The experimental
results confirm our theoretical analysis.
|
Rows vs Columns for Linear Systems of Equations - Randomized Kaczmarz or
Coordinate Descent? | This paper is about randomized iterative algorithms for solving a linear
system of equations $X \beta = y$ in different settings. Recent interest in the
topic was reignited when Strohmer and Vershynin (2009) proved the linear
convergence rate of a Randomized Kaczmarz (RK) algorithm that works on the rows
of $X$ (data points). Following that, Leventhal and Lewis (2010) proved the
linear convergence of a Randomized Coordinate Descent (RCD) algorithm that
works on the columns of $X$ (features). The aim of this paper is to simplify
our understanding of these two algorithms, establish the direct relationships
between them (though RK is often compared to Stochastic Gradient Descent), and
examine the algorithmic commonalities or tradeoffs involved with working on
rows or columns. We also discuss Kernel Ridge Regression and present a
Kaczmarz-style algorithm that works on data points and having the advantage of
solving the problem without ever storing or forming the Gram matrix, one of the
recognized problems encountered when scaling kernelized methods.
|
Semi-Supervised Learning with Deep Generative Models | The ever-increasing size of modern data sets combined with the difficulty of
obtaining label information has made semi-supervised learning one of the
problems of significant practical importance in modern data analysis. We
revisit the approach to semi-supervised learning with generative models and
develop new models that allow for effective generalisation from small labelled
data sets to large unlabelled ones. Generative approaches have thus far been
either inflexible, inefficient or non-scalable. We show that deep generative
models and approximate Bayesian inference exploiting recent advances in
variational methods can be used to provide significant improvements, making
generative approaches highly competitive for semi-supervised learning.
|
Towards A Deeper Geometric, Analytic and Algorithmic Understanding of
Margins | Given a matrix $A$, a linear feasibility problem (of which linear
classification is a special case) aims to find a solution to a primal problem
$w: A^Tw > \textbf{0}$ or a certificate for the dual problem which is a
probability distribution $p: Ap = \textbf{0}$. Inspired by the continued
importance of "large-margin classifiers" in machine learning, this paper
studies a condition measure of $A$ called its \textit{margin} that determines
the difficulty of both the above problems. To aid geometrical intuition, we
first establish new characterizations of the margin in terms of relevant balls,
cones and hulls. Our second contribution is analytical, where we present
generalizations of Gordan's theorem, and variants of Hoffman's theorems, both
using margins. We end by proving some new results on a classical iterative
scheme, the Perceptron, whose convergence rates famously depends on the margin.
Our results are relevant for a deeper understanding of margin-based learning
and proving convergence rates of iterative schemes, apart from providing a
unifying perspective on this vast topic.
|
Predicting the Future Behavior of a Time-Varying Probability
Distribution | We study the problem of predicting the future, though only in the
probabilistic sense of estimating a future state of a time-varying probability
distribution. This is not only an interesting academic problem, but solving
this extrapolation problem also has many practical application, e.g. for
training classifiers that have to operate under time-varying conditions. Our
main contribution is a method for predicting the next step of the time-varying
distribution from a given sequence of sample sets from earlier time steps. For
this we rely on two recent machine learning techniques: embedding probability
distributions into a reproducing kernel Hilbert space, and learning operators
by vector-valued regression. We illustrate the working principles and the
practical usefulness of our method by experiments on synthetic and real data.
We also highlight an exemplary application: training a classifier in a domain
adaptation setting without having access to examples from the test time
distribution at training time.
|
Spectral Ranking using Seriation | We describe a seriation algorithm for ranking a set of items given pairwise
comparisons between these items. Intuitively, the algorithm assigns similar
rankings to items that compare similarly with all others. It does so by
constructing a similarity matrix from pairwise comparisons, using seriation
methods to reorder this matrix and construct a ranking. We first show that this
spectral seriation algorithm recovers the true ranking when all pairwise
comparisons are observed and consistent with a total order. We then show that
ranking reconstruction is still exact when some pairwise comparisons are
corrupted or missing, and that seriation based spectral ranking is more robust
to noise than classical scoring methods. Finally, we bound the ranking error
when only a random subset of the comparions are observed. An additional benefit
of the seriation formulation is that it allows us to solve semi-supervised
ranking problems. Experiments on both synthetic and real datasets demonstrate
that seriation based spectral ranking achieves competitive and in some cases
superior performance compared to classical ranking methods.
|
Noise-adaptive Margin-based Active Learning and Lower Bounds under
Tsybakov Noise Condition | We present a simple noise-robust margin-based active learning algorithm to
find homogeneous (passing the origin) linear separators and analyze its error
convergence when labels are corrupted by noise. We show that when the imposed
noise satisfies the Tsybakov low noise condition (Mammen, Tsybakov, and others
1999; Tsybakov 2004) the algorithm is able to adapt to unknown level of noise
and achieves optimal statistical rate up to poly-logarithmic factors. We also
derive lower bounds for margin based active learning algorithms under Tsybakov
noise conditions (TNC) for the membership query synthesis scenario (Angluin
1988). Our result implies lower bounds for the stream based selective sampling
scenario (Cohn 1990) under TNC for some fairly simple data distributions. Quite
surprisingly, we show that the sample complexity cannot be improved even if the
underlying data distribution is as simple as the uniform distribution on the
unit ball. Our proof involves the construction of a well separated hypothesis
set on the d-dimensional unit ball along with carefully designed label
distributions for the Tsybakov noise condition. Our analysis might provide
insights for other forms of lower bounds as well.
|
Learning computationally efficient dictionaries and their implementation
as fast transforms | Dictionary learning is a branch of signal processing and machine learning
that aims at finding a frame (called dictionary) in which some training data
admits a sparse representation. The sparser the representation, the better the
dictionary. The resulting dictionary is in general a dense matrix, and its
manipulation can be computationally costly both at the learning stage and later
in the usage of this dictionary, for tasks such as sparse coding. Dictionary
learning is thus limited to relatively small-scale problems. In this paper,
inspired by usual fast transforms, we consider a general dictionary structure
that allows cheaper manipulation, and propose an algorithm to learn such
dictionaries --and their fast implementation-- over training data. The approach
is demonstrated experimentally with the factorization of the Hadamard matrix
and with synthetic dictionary learning experiments.
|
Playing with Duality: An Overview of Recent Primal-Dual Approaches for
Solving Large-Scale Optimization Problems | Optimization methods are at the core of many problems in signal/image
processing, computer vision, and machine learning. For a long time, it has been
recognized that looking at the dual of an optimization problem may drastically
simplify its solution. Deriving efficient strategies which jointly brings into
play the primal and the dual problems is however a more recent idea which has
generated many important new contributions in the last years. These novel
developments are grounded on recent advances in convex analysis, discrete
optimization, parallel processing, and non-smooth optimization with emphasis on
sparsity issues. In this paper, we aim at presenting the principles of
primal-dual approaches, while giving an overview of numerical methods which
have been proposed in different contexts. We show the benefits which can be
drawn from primal-dual algorithms both for solving large-scale convex
optimization problems and discrete ones, and we provide various application
examples to illustrate their usefulness.
|
An Open Source Pattern Recognition Toolbox for MATLAB | Pattern recognition and machine learning are becoming integral parts of
algorithms in a wide range of applications. Different algorithms and approaches
for machine learning include different tradeoffs between performance and
computation, so during algorithm development it is often necessary to explore a
variety of different approaches to a given task. A toolbox with a unified
framework across multiple pattern recognition techniques enables algorithm
developers the ability to rapidly evaluate different choices prior to
deployment. MATLAB is a widely used environment for algorithm development and
prototyping, and although several MATLAB toolboxes for pattern recognition are
currently available these are either incomplete, expensive, or restrictively
licensed. In this work we describe a MATLAB toolbox for pattern recognition and
machine learning known as the PRT (Pattern Recognition Toolbox), licensed under
the permissive MIT license. The PRT includes many popular techniques for data
preprocessing, supervised learning, clustering, regression and feature
selection, as well as a methodology for combining these components using a
simple, uniform syntax. The resulting algorithms can be evaluated using
cross-validation and a variety of scoring metrics to ensure robust performance
when the algorithm is deployed. This paper presents an overview of the PRT as
well as an example of usage on Fisher's Iris dataset.
|
From conformal to probabilistic prediction | This paper proposes a new method of probabilistic prediction, which is based
on conformal prediction. The method is applied to the standard USPS data set
and gives encouraging results.
|
PAC-Bayes Analysis of Multi-view Learning | This paper presents eight PAC-Bayes bounds to analyze the generalization
performance of multi-view classifiers. These bounds adopt data dependent
Gaussian priors which emphasize classifiers with high view agreements. The
center of the prior for the first two bounds is the origin, while the center of
the prior for the third and fourth bounds is given by a data dependent vector.
An important technique to obtain these bounds is two derived logarithmic
determinant inequalities whose difference lies in whether the dimensionality of
data is involved. The centers of the fifth and sixth bounds are calculated on a
separate subset of the training set. The last two bounds use unlabeled data to
represent view agreements and are thus applicable to semi-supervised multi-view
learning. We evaluate all the presented multi-view PAC-Bayes bounds on
benchmark data and compare them with previous single-view PAC-Bayes bounds. The
usefulness and performance of the multi-view bounds are discussed.
|
On semidefinite relaxations for the block model | The stochastic block model (SBM) is a popular tool for community detection in
networks, but fitting it by maximum likelihood (MLE) involves a computationally
infeasible optimization problem. We propose a new semidefinite programming
(SDP) solution to the problem of fitting the SBM, derived as a relaxation of
the MLE. We put ours and previously proposed SDPs in a unified framework, as
relaxations of the MLE over various sub-classes of the SBM, revealing a
connection to sparse PCA. Our main relaxation, which we call SDP-1, is tighter
than other recently proposed SDP relaxations, and thus previously established
theoretical guarantees carry over. However, we show that SDP-1 exactly recovers
true communities over a wider class of SBMs than those covered by current
results. In particular, the assumption of strong assortativity of the SBM,
implicit in consistency conditions for previously proposed SDPs, can be relaxed
to weak assortativity for our approach, thus significantly broadening the class
of SBMs covered by the consistency results. We also show that strong
assortativity is indeed a necessary condition for exact recovery for previously
proposed SDP approaches and not an artifact of the proofs. Our analysis of SDPs
is based on primal-dual witness constructions, which provides some insight into
the nature of the solutions of various SDPs. We show how to combine features
from SDP-1 and already available SDPs to achieve the most flexibility in terms
of both assortativity and block-size constraints, as our relaxation has the
tendency to produce communities of similar sizes. This tendency makes it the
ideal tool for fitting network histograms, a method gaining popularity in the
graphon estimation literature, as we illustrate on an example of a social
networks of dolphins. We also provide empirical evidence that SDPs outperform
spectral methods for fitting SBMs with a large number of blocks.
|
Constant Factor Approximation for Balanced Cut in the PIE model | We propose and study a new semi-random semi-adversarial model for Balanced
Cut, a planted model with permutation-invariant random edges (PIE). Our model
is much more general than planted models considered previously. Consider a set
of vertices V partitioned into two clusters $L$ and $R$ of equal size. Let $G$
be an arbitrary graph on $V$ with no edges between $L$ and $R$. Let
$E_{random}$ be a set of edges sampled from an arbitrary permutation-invariant
distribution (a distribution that is invariant under permutation of vertices in
$L$ and in $R$). Then we say that $G + E_{random}$ is a graph with
permutation-invariant random edges.
We present an approximation algorithm for the Balanced Cut problem that finds
a balanced cut of cost $O(|E_{random}|) + n \text{polylog}(n)$ in this model.
In the regime when $|E_{random}| = \Omega(n \text{polylog}(n))$, this is a
constant factor approximation with respect to the cost of the planted cut.
|
Correlation Clustering with Noisy Partial Information | In this paper, we propose and study a semi-random model for the Correlation
Clustering problem on arbitrary graphs G. We give two approximation algorithms
for Correlation Clustering instances from this model. The first algorithm finds
a solution of value $(1+ \delta) optcost + O_{\delta}(n\log^3 n)$ with high
probability, where $optcost$ is the value of the optimal solution (for every
$\delta > 0$). The second algorithm finds the ground truth clustering with an
arbitrarily small classification error $\eta$ (under some additional
assumptions on the instance).
|
SPSD Matrix Approximation vis Column Selection: Theories, Algorithms,
and Extensions | Symmetric positive semidefinite (SPSD) matrix approximation is an important
problem with applications in kernel methods. However, existing SPSD matrix
approximation methods such as the Nystr\"om method only have weak error bounds.
In this paper we conduct in-depth studies of an SPSD matrix approximation model
and establish strong relative-error bounds. We call it the prototype model for
it has more efficient and effective extensions, and some of its extensions have
high scalability. Though the prototype model itself is not suitable for
large-scale data, it is still useful to study its properties, on which the
analysis of its extensions relies.
This paper offers novel theoretical analysis, efficient algorithms, and a
highly accurate extension. First, we establish a lower error bound for the
prototype model and improve the error bound of an existing column selection
algorithm to match the lower bound. In this way, we obtain the first optimal
column selection algorithm for the prototype model. We also prove that the
prototype model is exact under certain conditions. Second, we develop a simple
column selection algorithm with a provable error bound. Third, we propose a
so-called spectral shifting model to make the approximation more accurate when
the eigenvalues of the matrix decay slowly, and the improvement is
theoretically quantified. The spectral shifting method can also be applied to
improve other SPSD matrix approximation models.
|
Deep Fragment Embeddings for Bidirectional Image Sentence Mapping | We introduce a model for bidirectional retrieval of images and sentences
through a multi-modal embedding of visual and natural language data. Unlike
previous models that directly map images or sentences into a common embedding
space, our model works on a finer level and embeds fragments of images
(objects) and fragments of sentences (typed dependency tree relations) into a
common space. In addition to a ranking objective seen in previous work, this
allows us to add a new fragment alignment objective that learns to directly
associate these fragments across modalities. Extensive experimental evaluation
shows that reasoning on both the global level of images and sentences and the
finer level of their respective fragments significantly improves performance on
image-sentence retrieval tasks. Additionally, our model provides interpretable
predictions since the inferred inter-modal fragment alignment is explicit.
|
On the Maximum Entropy Property of the First-Order Stable Spline Kernel
and its Implications | A new nonparametric approach for system identification has been recently
proposed where the impulse response is seen as the realization of a zero--mean
Gaussian process whose covariance, the so--called stable spline kernel,
guarantees that the impulse response is almost surely stable. Maximum entropy
properties of the stable spline kernel have been pointed out in the literature.
In this paper we provide an independent proof that relies on the theory of
matrix extension problems in the graphical model literature and leads to a
closed form expression for the inverse of the first order stable spline kernel
as well as to a new factorization in the form $UWU^\top$ with $U$ upper
triangular and $W$ diagonal. Interestingly, all first--order stable spline
kernels share the same factor $U$ and $W$ admits a closed form representation
in terms of the kernel hyperparameter, making the factorization computationally
inexpensive. Maximum likelihood properties of the stable spline kernel are also
highlighted. These results can be applied both to improve the stability and to
reduce the computational complexity associated with the computation of stable
spline estimators.
|
Convex Optimization Learning of Faithful Euclidean Distance
Representations in Nonlinear Dimensionality Reduction | Classical multidimensional scaling only works well when the noisy distances
observed in a high dimensional space can be faithfully represented by Euclidean
distances in a low dimensional space. Advanced models such as Maximum Variance
Unfolding (MVU) and Minimum Volume Embedding (MVE) use Semi-Definite
Programming (SDP) to reconstruct such faithful representations. While those SDP
models are capable of producing high quality configuration numerically, they
suffer two major drawbacks. One is that there exist no theoretically guaranteed
bounds on the quality of the configuration. The other is that they are slow in
computation when the data points are beyond moderate size. In this paper, we
propose a convex optimization model of Euclidean distance matrices. We
establish a non-asymptotic error bound for the random graph model with
sub-Gaussian noise, and prove that our model produces a matrix estimator of
high accuracy when the order of the uniform sample size is roughly the degree
of freedom of a low-rank matrix up to a logarithmic factor. Our results
partially explain why MVU and MVE often work well. Moreover, we develop a fast
inexact accelerated proximal gradient method. Numerical experiments show that
the model can produce configurations of high quality on large data points that
the SDP approach would struggle to cope with.
|
Divide-and-Conquer Learning by Anchoring a Conical Hull | We reduce a broad class of machine learning problems, usually addressed by EM
or sampling, to the problem of finding the $k$ extremal rays spanning the
conical hull of a data point set. These $k$ "anchors" lead to a global solution
and a more interpretable model that can even outperform EM and sampling on
generalization error. To find the $k$ anchors, we propose a novel
divide-and-conquer learning scheme "DCA" that distributes the problem to
$\mathcal O(k\log k)$ same-type sub-problems on different low-D random
hyperplanes, each can be solved by any solver. For the 2D sub-problem, we
present a non-iterative solver that only needs to compute an array of cosine
values and its max/min entries. DCA also provides a faster subroutine for other
methods to check whether a point is covered in a conical hull, which improves
algorithm design in multiple dimensions and brings significant speedup to
learning. We apply our method to GMM, HMM, LDA, NMF and subspace clustering,
then show its competitive performance and scalability over other methods on
rich datasets.
|
Multi-utility Learning: Structured-output Learning with Multiple
Annotation-specific Loss Functions | Structured-output learning is a challenging problem; particularly so because
of the difficulty in obtaining large datasets of fully labelled instances for
training. In this paper we try to overcome this difficulty by presenting a
multi-utility learning framework for structured prediction that can learn from
training instances with different forms of supervision. We propose a unified
technique for inferring the loss functions most suitable for quantifying the
consistency of solutions with the given weak annotation. We demonstrate the
effectiveness of our framework on the challenging semantic image segmentation
problem for which a wide variety of annotations can be used. For instance, the
popular training datasets for semantic segmentation are composed of images with
hard-to-generate full pixel labellings, as well as images with easy-to-obtain
weak annotations, such as bounding boxes around objects, or image-level labels
that specify which object categories are present in an image. Experimental
evaluation shows that the use of annotation-specific loss functions
dramatically improves segmentation accuracy compared to the baseline system
where only one type of weak annotation is used.
|
Reinforcement and Imitation Learning via Interactive No-Regret Learning | Recent work has demonstrated that problems-- particularly imitation learning
and structured prediction-- where a learner's predictions influence the
input-distribution it is tested on can be naturally addressed by an interactive
approach and analyzed using no-regret online learning. These approaches to
imitation learning, however, neither require nor benefit from information about
the cost of actions. We extend existing results in two directions: first, we
develop an interactive imitation learning approach that leverages cost
information; second, we extend the technique to address reinforcement learning.
The results provide theoretical support to the commonly observed successes of
online approximate policy iteration. Our approach suggests a broad new family
of algorithms and provides a unifying view of existing techniques for imitation
and reinforcement learning.
|
Stationary Mixing Bandits | We study the bandit problem where arms are associated with stationary
phi-mixing processes and where rewards are therefore dependent: the question
that arises from this setting is that of recovering some independence by
ignoring the value of some rewards. As we shall see, the bandit problem we
tackle requires us to address the exploration/exploitation/independence
trade-off. To do so, we provide a UCB strategy together with a general regret
analysis for the case where the size of the independence blocks (the ignored
rewards) is fixed and we go a step beyond by providing an algorithm that is
able to compute the size of the independence blocks from the data. Finally, we
give an analysis of our bandit problem in the restless case, i.e., in the
situation where the time counters for all mixing processes simultaneously
evolve.
|
From Black-Scholes to Online Learning: Dynamic Hedging under Adversarial
Environments | We consider a non-stochastic online learning approach to price financial
options by modeling the market dynamic as a repeated game between the nature
(adversary) and the investor. We demonstrate that such framework yields
analogous structure as the Black-Scholes model, the widely popular option
pricing model in stochastic finance, for both European and American options
with convex payoffs. In the case of non-convex options, we construct
approximate pricing algorithms, and demonstrate that their efficiency can be
analyzed through the introduction of an artificial probability measure, in
parallel to the so-called risk-neutral measure in the finance literature, even
though our framework is completely adversarial. Continuous-time convergence
results and extensions to incorporate price jumps are also presented.
|
Improved Frame Level Features and SVM Supervectors Approach for the
Recogniton of Emotional States from Speech: Application to categorical and
dimensional states | The purpose of speech emotion recognition system is to classify speakers
utterances into different emotional states such as disgust, boredom, sadness,
neutral and happiness. Speech features that are commonly used in speech emotion
recognition rely on global utterance level prosodic features. In our work, we
evaluate the impact of frame level feature extraction. The speech samples are
from Berlin emotional database and the features extracted from these utterances
are energy, different variant of mel frequency cepstrum coefficients, velocity
and acceleration features.
|
Mining Recurrent Concepts in Data Streams using the Discrete Fourier
Transform | In this research we address the problem of capturing recurring concepts in a
data stream environment. Recurrence capture enables the re-use of previously
learned classifiers without the need for re-learning while providing for better
accuracy during the concept recurrence interval. We capture concepts by
applying the Discrete Fourier Transform (DFT) to Decision Tree classifiers to
obtain highly compressed versions of the trees at concept drift points in the
stream and store such trees in a repository for future use. Our empirical
results on real world and synthetic data exhibiting varying degrees of
recurrence show that the Fourier compressed trees are more robust to noise and
are able to capture recurring concepts with higher precision than a meta
learning approach that chooses to re-use classifiers in their originally
occurring form.
|
Generalized Mixability via Entropic Duality | Mixability is a property of a loss which characterizes when fast convergence
is possible in the game of prediction with expert advice. We show that a key
property of mixability generalizes, and the exp and log operations present in
the usual theory are not as special as one might have thought. In doing this we
introduce a more general notion of $\Phi$-mixability where $\Phi$ is a general
entropy (\ie, any convex function on probabilities). We show how a property
shared by the convex dual of any such entropy yields a natural algorithm (the
minimizer of a regret bound) which, analogous to the classical aggregating
algorithm, is guaranteed a constant regret when used with $\Phi$-mixable
losses. We characterize precisely which $\Phi$ have $\Phi$-mixable losses and
put forward a number of conjectures about the optimality and relationships
between different choices of entropy.
|
Fast, Robust and Non-convex Subspace Recovery | This work presents a fast and non-convex algorithm for robust subspace
recovery. The data sets considered include inliers drawn around a
low-dimensional subspace of a higher dimensional ambient space, and a possibly
large portion of outliers that do not lie nearby this subspace. The proposed
algorithm, which we refer to as Fast Median Subspace (FMS), is designed to
robustly determine the underlying subspace of such data sets, while having
lower computational complexity than existing methods. We prove convergence of
the FMS iterates to a stationary point. Further, under a special model of data,
FMS converges to a point which is near to the global minimum with overwhelming
probability. Under this model, we show that the iteration complexity is
globally bounded and locally $r$-linear. The latter theorem holds for any fixed
fraction of outliers (less than 1) and any fixed positive distance between the
limit point and the global minimum. Numerical experiments on synthetic and real
data demonstrate its competitive speed and accuracy.
|
Composite Likelihood Estimation for Restricted Boltzmann machines | Learning the parameters of graphical models using the maximum likelihood
estimation is generally hard which requires an approximation. Maximum composite
likelihood estimations are statistical approximations of the maximum likelihood
estimation which are higher-order generalizations of the maximum
pseudo-likelihood estimation. In this paper, we propose a composite likelihood
method and investigate its property. Furthermore, we apply our composite
likelihood method to restricted Boltzmann machines.
|
Combining predictions from linear models when training and test inputs
differ | Methods for combining predictions from different models in a supervised
learning setting must somehow estimate/predict the quality of a model's
predictions at unknown future inputs. Many of these methods (often implicitly)
make the assumption that the test inputs are identical to the training inputs,
which is seldom reasonable. By failing to take into account that prediction
will generally be harder for test inputs that did not occur in the training
set, this leads to the selection of too complex models. Based on a novel,
unbiased expression for KL divergence, we propose XAIC and its special case
FAIC as versions of AIC intended for prediction that use different degrees of
knowledge of the test inputs. Both methods substantially differ from and may
outperform all the known versions of AIC even when the training and test inputs
are iid, and are especially useful for deterministic inputs and under covariate
shift. Our experiments on linear models suggest that if the test and training
inputs differ substantially, then XAIC and FAIC predictively outperform AIC,
BIC and several other methods including Bayesian model averaging.
|
Recurrent Models of Visual Attention | Applying convolutional neural networks to large images is computationally
expensive because the amount of computation scales linearly with the number of
image pixels. We present a novel recurrent neural network model that is capable
of extracting information from an image or video by adaptively selecting a
sequence of regions or locations and only processing the selected regions at
high resolution. Like convolutional neural networks, the proposed model has a
degree of translation invariance built-in, but the amount of computation it
performs can be controlled independently of the input image size. While the
model is non-differentiable, it can be trained using reinforcement learning
methods to learn task-specific policies. We evaluate our model on several image
classification tasks, where it significantly outperforms a convolutional neural
network baseline on cluttered images, and on a dynamic visual control problem,
where it learns to track a simple object without an explicit training signal
for doing so.
|
Scalable Topical Phrase Mining from Text Corpora | While most topic modeling algorithms model text corpora with unigrams, human
interpretation often relies on inherent grouping of terms into phrases. As
such, we consider the problem of discovering topical phrases of mixed lengths.
Existing work either performs post processing to the inference results of
unigram-based topic models, or utilizes complex n-gram-discovery topic models.
These methods generally produce low-quality topical phrases or suffer from poor
scalability on even moderately-sized datasets. We propose a different approach
that is both computationally efficient and effective. Our solution combines a
novel phrase mining framework to segment a document into single and multi-word
phrases, and a new topic model that operates on the induced document partition.
Our approach discovers high quality topical phrases with negligible extra cost
to the bag-of-words topic model in a variety of datasets including research
publication titles, abstracts, reviews, and news articles.
|
Further heuristics for $k$-means: The merge-and-split heuristic and the
$(k,l)$-means | Finding the optimal $k$-means clustering is NP-hard in general and many
heuristics have been designed for minimizing monotonically the $k$-means
objective. We first show how to extend Lloyd's batched relocation heuristic and
Hartigan's single-point relocation heuristic to take into account empty-cluster
and single-point cluster events, respectively. Those events tend to
increasingly occur when $k$ or $d$ increases, or when performing several
restarts. First, we show that those special events are a blessing because they
allow to partially re-seed some cluster centers while further minimizing the
$k$-means objective function. Second, we describe a novel heuristic,
merge-and-split $k$-means, that consists in merging two clusters and splitting
this merged cluster again with two new centers provided it improves the
$k$-means objective. This novel heuristic can improve Hartigan's $k$-means when
it has converged to a local minimum. We show empirically that this
merge-and-split $k$-means improves over the Hartigan's heuristic which is the
{\em de facto} method of choice. Finally, we propose the $(k,l)$-means
objective that generalizes the $k$-means objective by associating the data
points to their $l$ closest cluster centers, and show how to either directly
convert or iteratively relax the $(k,l)$-means into a $k$-means in order to
reach better local minima.
|
Incremental Clustering: The Case for Extra Clusters | The explosion in the amount of data available for analysis often necessitates
a transition from batch to incremental clustering methods, which process one
element at a time and typically store only a small subset of the data. In this
paper, we initiate the formal analysis of incremental clustering methods
focusing on the types of cluster structure that they are able to detect. We
find that the incremental setting is strictly weaker than the batch model,
proving that a fundamental class of cluster structures that can readily be
detected in the batch setting is impossible to identify using any incremental
method. Furthermore, we show how the limitations of incremental clustering can
be overcome by allowing additional clusters.
|
On the Convergence Rate of Decomposable Submodular Function Minimization | Submodular functions describe a variety of discrete problems in machine
learning, signal processing, and computer vision. However, minimizing
submodular functions poses a number of algorithmic challenges. Recent work
introduced an easy-to-use, parallelizable algorithm for minimizing submodular
functions that decompose as the sum of "simple" submodular functions.
Empirically, this algorithm performs extremely well, but no theoretical
analysis was given. In this paper, we show that the algorithm converges
linearly, and we provide upper and lower bounds on the rate of convergence. Our
proof relies on the geometry of submodular polyhedra and draws on results from
spectral graph theory.
|
Weakly-supervised Discovery of Visual Pattern Configurations | The increasing prominence of weakly labeled data nurtures a growing demand
for object detection methods that can cope with minimal supervision. We propose
an approach that automatically identifies discriminative configurations of
visual patterns that are characteristic of a given object class. We formulate
the problem as a constrained submodular optimization problem and demonstrate
the benefits of the discovered configurations in remedying mislocalizations and
finding informative positive and negative training examples. Together, these
lead to state-of-the-art weakly-supervised detection results on the challenging
PASCAL VOC dataset.
|
Support vector machine classification of dimensionally reduced
structural MRI images for dementia | We classify very-mild to moderate dementia in patients (CDR ranging from 0 to
2) using a support vector machine classifier acting on dimensionally reduced
feature set derived from MRI brain scans of the 416 subjects available in the
OASIS-Brains dataset. We use image segmentation and principal component
analysis to reduce the dimensionality of the data. Our resulting feature set
contains 11 features for each subject. Performance of the classifiers is
evaluated using 10-fold cross-validation. Using linear and (gaussian) kernels,
we obtain a training classification accuracy of 86.4% (90.1%), test accuracy of
85.0% (85.7%), test precision of 68.7% (68.5%), test recall of 68.0% (74.0%),
and test Matthews correlation coefficient of 0.594 (0.616).
|
A scaled gradient projection method for Bayesian learning in dynamical
systems | A crucial task in system identification problems is the selection of the most
appropriate model class, and is classically addressed resorting to
cross-validation or using asymptotic arguments. As recently suggested in the
literature, this can be addressed in a Bayesian framework, where model
complexity is regulated by few hyperparameters, which can be estimated via
marginal likelihood maximization. It is thus of primary importance to design
effective optimization methods to solve the corresponding optimization problem.
If the unknown impulse response is modeled as a Gaussian process with a
suitable kernel, the maximization of the marginal likelihood leads to a
challenging nonconvex optimization problem, which requires a stable and
effective solution strategy. In this paper we address this problem by means of
a scaled gradient projection algorithm, in which the scaling matrix and the
steplength parameter play a crucial role to provide a meaning solution in a
computational time comparable with second order methods. In particular, we
propose both a generalization of the split gradient approach to design the
scaling matrix in the presence of box constraints, and an effective
implementation of the gradient and objective function. The extensive numerical
experiments carried out on several test problems show that our method is very
effective in providing in few tenths of a second solutions of the problems with
accuracy comparable with state-of-the-art approaches. Moreover, the flexibility
of the proposed strategy makes it easily adaptable to a wider range of problems
arising in different areas of machine learning, signal processing and system
identification.
|
When is it Better to Compare than to Score? | When eliciting judgements from humans for an unknown quantity, one often has
the choice of making direct-scoring (cardinal) or comparative (ordinal)
measurements. In this paper we study the relative merits of either choice,
providing empirical and theoretical guidelines for the selection of a
measurement scheme. We provide empirical evidence based on experiments on
Amazon Mechanical Turk that in a variety of tasks, (pairwise-comparative)
ordinal measurements have lower per sample noise and are typically faster to
elicit than cardinal ones. Ordinal measurements however typically provide less
information. We then consider the popular Thurstone and Bradley-Terry-Luce
(BTL) models for ordinal measurements and characterize the minimax error rates
for estimating the unknown quantity. We compare these minimax error rates to
those under cardinal measurement models and quantify for what noise levels
ordinal measurements are better. Finally, we revisit the data collected from
our experiments and show that fitting these models confirms this prediction:
for tasks where the noise in ordinal measurements is sufficiently low, the
ordinal approach results in smaller errors in the estimation.
|
Active Learning and Best-Response Dynamics | We examine an important setting for engineered systems in which low-power
distributed sensors are each making highly noisy measurements of some unknown
target function. A center wants to accurately learn this function by querying a
small number of sensors, which ordinarily would be impossible due to the high
noise rate. The question we address is whether local communication among
sensors, together with natural best-response dynamics in an
appropriately-defined game, can denoise the system without destroying the true
signal and allow the center to succeed from only a small number of active
queries. By using techniques from game theory and empirical processes, we prove
positive (and negative) results on the denoising power of several natural
dynamics. We then show experimentally that when combined with recent agnostic
active learning algorithms, this process can achieve low error from very few
queries, performing substantially better than active or passive learning
without these denoising dynamics as well as passive learning with denoising.
|
Causality Networks | While correlation measures are used to discern statistical relationships
between observed variables in almost all branches of data-driven scientific
inquiry, what we are really interested in is the existence of causal
dependence. Designing an efficient causality test, that may be carried out in
the absence of restrictive pre-suppositions on the underlying dynamical
structure of the data at hand, is non-trivial. Nevertheless, ability to
computationally infer statistical prima facie evidence of causal dependence may
yield a far more discriminative tool for data analysis compared to the
calculation of simple correlations. In the present work, we present a new
non-parametric test of Granger causality for quantized or symbolic data streams
generated by ergodic stationary sources. In contrast to state-of-art binary
tests, our approach makes precise and computes the degree of causal dependence
between data streams, without making any restrictive assumptions, linearity or
otherwise. Additionally, without any a priori imposition of specific dynamical
structure, we infer explicit generative models of causal cross-dependence,
which may be then used for prediction. These explicit models are represented as
generalized probabilistic automata, referred to crossed automata, and are shown
to be sufficient to capture a fairly general class of causal dependence. The
proposed algorithms are computationally efficient in the PAC sense; $i.e.$, we
find good models of cross-dependence with high probability, with polynomial
run-times and sample complexities. The theoretical results are applied to
weekly search-frequency data from Google Trends API for a chosen set of
socially "charged" keywords. The causality network inferred from this dataset
reveals, quite expectedly, the causal importance of certain keywords. It is
also illustrated that correlation analysis fails to gather such insight.
|
Mass-Univariate Hypothesis Testing on MEEG Data using Cross-Validation | Recent advances in statistical theory, together with advances in the
computational power of computers, provide alternative methods to do
mass-univariate hypothesis testing in which a large number of univariate tests,
can be properly used to compare MEEG data at a large number of time-frequency
points and scalp locations. One of the major problematic aspects of this kind
of mass-univariate analysis is due to high number of accomplished hypothesis
tests. Hence procedures that remove or alleviate the increased probability of
false discoveries are crucial for this type of analysis. Here, I propose a new
method for mass-univariate analysis of MEEG data based on cross-validation
scheme. In this method, I suggest a hierarchical classification procedure under
k-fold cross-validation to detect which sensors at which time-bin and which
frequency-bin contributes in discriminating between two different stimuli or
tasks. To achieve this goal, a new feature extraction method based on the
discrete cosine transform (DCT) employed to get maximum advantage of all three
data dimensions. Employing cross-validation and hierarchy architecture
alongside the DCT feature space makes this method more reliable and at the same
time enough sensitive to detect the narrow effects in brain activities.
|
Online learning in MDPs with side information | We study online learning of finite Markov decision process (MDP) problems
when a side information vector is available. The problem is motivated by
applications such as clinical trials, recommendation systems, etc. Such
applications have an episodic structure, where each episode corresponds to a
patient/customer. Our objective is to compete with the optimal dynamic policy
that can take side information into account.
We propose a computationally efficient algorithm and show that its regret is
at most $O(\sqrt{T})$, where $T$ is the number of rounds. To best of our
knowledge, this is the first regret bound for this setting.
|
Discriminative Unsupervised Feature Learning with Exemplar Convolutional
Neural Networks | Deep convolutional networks have proven to be very successful in learning
task specific features that allow for unprecedented performance on various
computer vision tasks. Training of such networks follows mostly the supervised
learning paradigm, where sufficiently many input-output pairs are required for
training. Acquisition of large training sets is one of the key challenges, when
approaching a new task. In this paper, we aim for generic feature learning and
present an approach for training a convolutional network using only unlabeled
data. To this end, we train the network to discriminate between a set of
surrogate classes. Each surrogate class is formed by applying a variety of
transformations to a randomly sampled 'seed' image patch. In contrast to
supervised network training, the resulting feature representation is not class
specific. It rather provides robustness to the transformations that have been
applied during training. This generic feature representation allows for
classification results that outperform the state of the art for unsupervised
learning on several popular datasets (STL-10, CIFAR-10, Caltech-101,
Caltech-256). While such generic features cannot compete with class specific
features from supervised training on a classification task, we show that they
are advantageous on geometric matching problems, where they also outperform the
SIFT descriptor.
|
A Concise Information-Theoretic Derivation of the Baum-Welch algorithm | We derive the Baum-Welch algorithm for hidden Markov models (HMMs) through an
information-theoretical approach using cross-entropy instead of the Lagrange
multiplier approach which is universal in machine learning literature. The
proposed approach provides a more concise derivation of the Baum-Welch method
and naturally generalizes to multiple observations.
|
An Incentive Compatible Multi-Armed-Bandit Crowdsourcing Mechanism with
Quality Assurance | Consider a requester who wishes to crowdsource a series of identical binary
labeling tasks to a pool of workers so as to achieve an assured accuracy for
each task, in a cost optimal way. The workers are heterogeneous with unknown
but fixed qualities and their costs are private. The problem is to select for
each task an optimal subset of workers so that the outcome obtained from the
selected workers guarantees a target accuracy level. The problem is a
challenging one even in a non strategic setting since the accuracy of
aggregated label depends on unknown qualities. We develop a novel multi-armed
bandit (MAB) mechanism for solving this problem. First, we propose a framework,
Assured Accuracy Bandit (AAB), which leads to an MAB algorithm, Constrained
Confidence Bound for a Non Strategic setting (CCB-NS). We derive an upper bound
on the number of time steps the algorithm chooses a sub-optimal set that
depends on the target accuracy level and true qualities. A more challenging
situation arises when the requester not only has to learn the qualities of the
workers but also elicit their true costs. We modify the CCB-NS algorithm to
obtain an adaptive exploration separated algorithm which we call { \em
Constrained Confidence Bound for a Strategic setting (CCB-S)}. CCB-S algorithm
produces an ex-post monotone allocation rule and thus can be transformed into
an ex-post incentive compatible and ex-post individually rational mechanism
that learns the qualities of the workers and guarantees a given target accuracy
level in a cost optimal way. We provide a lower bound on the number of times
any algorithm should select a sub-optimal set and we see that the lower bound
matches our upper bound upto a constant factor. We provide insights on the
practical implementation of this framework through an illustrative example and
we show the efficacy of our algorithms through simulations.
|
Reconstructing subclonal composition and evolution from whole genome
sequencing of tumors | Tumors often contain multiple subpopulations of cancerous cells defined by
distinct somatic mutations. We describe a new method, PhyloWGS, that can be
applied to WGS data from one or more tumor samples to reconstruct complete
genotypes of these subpopulations based on variant allele frequencies (VAFs) of
point mutations and population frequencies of structural variations. We
introduce a principled phylogenic correction for VAFs in loci affected by copy
number alterations and we show that this correction greatly improves subclonal
reconstruction compared to existing methods.
|
On the Use of Different Feature Extraction Methods for Linear and Non
Linear kernels | The speech feature extraction has been a key focus in robust speech
recognition research; it significantly affects the recognition performance. In
this paper, we first study a set of different features extraction methods such
as linear predictive coding (LPC), mel frequency cepstral coefficient (MFCC)
and perceptual linear prediction (PLP) with several features normalization
techniques like rasta filtering and cepstral mean subtraction (CMS). Based on
this, a comparative evaluation of these features is performed on the task of
text independent speaker identification using a combination between gaussian
mixture models (GMM) and linear and non-linear kernels based on support vector
machine (SVM).
|
Stock Market Prediction from WSJ: Text Mining via Sparse Matrix
Factorization | We revisit the problem of predicting directional movements of stock prices
based on news articles: here our algorithm uses daily articles from The Wall
Street Journal to predict the closing stock prices on the same day. We propose
a unified latent space model to characterize the "co-movements" between stock
prices and news articles. Unlike many existing approaches, our new model is
able to simultaneously leverage the correlations: (a) among stock prices, (b)
among news articles, and (c) between stock prices and news articles. Thus, our
model is able to make daily predictions on more than 500 stocks (most of which
are not even mentioned in any news article) while having low complexity. We
carry out extensive backtesting on trading strategies based on our algorithm.
The result shows that our model has substantially better accuracy rate (55.7%)
compared to many widely used algorithms. The return (56%) and Sharpe ratio due
to a trading strategy based on our model are also much higher than baseline
indices.
|
Exponentially Increasing the Capacity-to-Computation Ratio for
Conditional Computation in Deep Learning | Many state-of-the-art results obtained with deep networks are achieved with
the largest models that could be trained, and if more computation power was
available, we might be able to exploit much larger datasets in order to improve
generalization ability. Whereas in learning algorithms such as decision trees
the ratio of capacity (e.g., the number of parameters) to computation is very
favorable (up to exponentially more parameters than computation), the ratio is
essentially 1 for deep neural networks. Conditional computation has been
proposed as a way to increase the capacity of a deep neural network without
increasing the amount of computation required, by activating some parameters
and computation "on-demand", on a per-example basis. In this note, we propose a
novel parametrization of weight matrices in neural networks which has the
potential to increase up to exponentially the ratio of the number of parameters
to computation. The proposed approach is based on turning on some parameters
(weight matrices) when specific bit patterns of hidden unit activations are
obtained. In order to better control for the overfitting that might result, we
propose a parametrization that is tree-structured, where each node of the tree
corresponds to a prefix of a sequence of sign bits, or gating units, associated
with hidden units.
|
Complexity Measures and Concept Learning | The nature of concept learning is a core question in cognitive science.
Theories must account for the relative difficulty of acquiring different
concepts by supervised learners. For a canonical set of six category types, two
distinct orderings of classification difficulty have been found. One ordering,
which we call paradigm-specific, occurs when adult human learners classify
objects with easily distinguishable characteristics such as size, shape, and
shading. The general order occurs in all other known cases: when adult humans
classify objects with characteristics that are not readily distinguished (e.g.,
brightness, saturation, hue); for children and monkeys; and when categorization
difficulty is extrapolated from errors in identification learning. The
paradigm-specific order was found to be predictable mathematically by measuring
the logical complexity of tasks, i.e., how concisely the solution can be
represented by logical rules.
However, logical complexity explains only the paradigm-specific order but not
the general order. Here we propose a new difficulty measurement, information
complexity, that calculates the amount of uncertainty remaining when a subset
of the dimensions are specified. This measurement is based on Shannon entropy.
We show that, when the metric extracts minimal uncertainties, this new
measurement predicts the paradigm-specific order for the canonical six category
types, and when the metric extracts average uncertainties, this new measurement
predicts the general order. Moreover, for learning category types beyond the
canonical six, we find that the minimal-uncertainty formulation correctly
predicts the paradigm-specific order as well or better than existing metrics
(Boolean complexity and GIST) in most cases.
|
Comparison of SVM Optimization Techniques in the Primal | This paper examines the efficacy of different optimization techniques in a
primal formulation of a support vector machine (SVM). Three main techniques are
compared. The dataset used to compare all three techniques was the Sentiment
Analysis on Movie Reviews dataset, from kaggle.com.
|
Efficient Learning in Large-Scale Combinatorial Semi-Bandits | A stochastic combinatorial semi-bandit is an online learning problem where at
each step a learning agent chooses a subset of ground items subject to
combinatorial constraints, and then observes stochastic weights of these items
and receives their sum as a payoff. In this paper, we consider efficient
learning in large-scale combinatorial semi-bandits with linear generalization,
and as a solution, propose two learning algorithms called Combinatorial Linear
Thompson Sampling (CombLinTS) and Combinatorial Linear UCB (CombLinUCB). Both
algorithms are computationally efficient as long as the offline version of the
combinatorial problem can be solved efficiently. We establish that CombLinTS
and CombLinUCB are also provably statistically efficient under reasonable
assumptions, by developing regret bounds that are independent of the problem
scale (number of items) and sublinear in time. We also evaluate CombLinTS on a
variety of problems with thousands of items. Our experiment results demonstrate
that CombLinTS is scalable, robust to the choice of algorithm parameters, and
significantly outperforms the best of our baselines.
|
Learning to Deblur | We describe a learning-based approach to blind image deconvolution. It uses a
deep layered architecture, parts of which are borrowed from recent work on
neural network learning, and parts of which incorporate computations that are
specific to image deconvolution. The system is trained end-to-end on a set of
artificially generated training examples, enabling competitive performance in
blind deconvolution, both with respect to quality and runtime.
|
Contrastive Feature Induction for Efficient Structure Learning of
Conditional Random Fields | Structure learning of Conditional Random Fields (CRFs) can be cast into an
L1-regularized optimization problem. To avoid optimizing over a fully linked
model, gain-based or gradient-based feature selection methods start from an
empty model and incrementally add top ranked features to it. However, for
high-dimensional problems like statistical relational learning, training time
of these incremental methods can be dominated by the cost of evaluating the
gain or gradient of a large collection of candidate features. In this study we
propose a fast feature evaluation algorithm called Contrastive Feature
Induction (CFI), which only evaluates a subset of features that involve both
variables with high signals (deviation from mean) and variables with high
errors (residue). We prove that the gradient of candidate features can be
represented solely as a function of signals and errors, and that CFI is an
efficient approximation of gradient-based evaluation methods. Experiments on
synthetic and real data sets show competitive learning speed and accuracy of
CFI on pairwise CRFs, compared to state-of-the-art structure learning methods
such as full optimization over all features, and Grafting.
|
Unimodal Bandits without Smoothness | We consider stochastic bandit problems with a continuous set of arms and
where the expected reward is a continuous and unimodal function of the arm. No
further assumption is made regarding the smoothness and the structure of the
expected reward function. For these problems, we propose the Stochastic
Pentachotomy (SP) algorithm, and derive finite-time upper bounds on its regret
and optimization error. In particular, we show that, for any expected reward
function $\mu$ that behaves as $\mu(x)=\mu(x^\star)-C|x-x^\star|^\xi$ locally
around its maximizer $x^\star$ for some $\xi, C>0$, the SP algorithm is
order-optimal. Namely its regret and optimization error scale as
$O(\sqrt{T\log(T)})$ and $O(\sqrt{\log(T)/T})$, respectively, when the time
horizon $T$ grows large. These scalings are achieved without the knowledge of
$\xi$ and $C$. Our algorithm is based on asymptotically optimal sequential
statistical tests used to successively trim an interval that contains the best
arm with high probability. To our knowledge, the SP algorithm constitutes the
first sequential arm selection rule that achieves a regret and optimization
error scaling as $O(\sqrt{T})$ and $O(1/\sqrt{T})$, respectively, up to a
logarithmic factor for non-smooth expected reward functions, as well as for
smooth functions with unknown smoothness.
|
Thompson Sampling for Learning Parameterized Markov Decision Processes | We consider reinforcement learning in parameterized Markov Decision Processes
(MDPs), where the parameterization may induce correlation across transition
probabilities or rewards. Consequently, observing a particular state transition
might yield useful information about other, unobserved, parts of the MDP. We
present a version of Thompson sampling for parameterized reinforcement learning
problems, and derive a frequentist regret bound for priors over general
parameter spaces. The result shows that the number of instants where suboptimal
actions are chosen scales logarithmically with time, with high probability. It
holds for prior distributions that put significant probability near the true
model, without any additional, specific closed-form structure such as conjugate
or product-form priors. The constant factor in the logarithmic scaling encodes
the information complexity of learning the MDP in terms of the Kullback-Leibler
geometry of the parameter space.
|
Theoretical Analysis of Bayesian Optimisation with Unknown Gaussian
Process Hyper-Parameters | Bayesian optimisation has gained great popularity as a tool for optimising
the parameters of machine learning algorithms and models. Somewhat ironically,
setting up the hyper-parameters of Bayesian optimisation methods is notoriously
hard. While reasonable practical solutions have been advanced, they can often
fail to find the best optima. Surprisingly, there is little theoretical
analysis of this crucial problem in the literature. To address this, we derive
a cumulative regret bound for Bayesian optimisation with Gaussian processes and
unknown kernel hyper-parameters in the stochastic setting. The bound, which
applies to the expected improvement acquisition function and sub-Gaussian
observation noise, provides us with guidelines on how to design hyper-parameter
estimation methods. A simple simulation demonstrates the importance of
following these guidelines.
|
Building DNN Acoustic Models for Large Vocabulary Speech Recognition | Deep neural networks (DNNs) are now a central component of nearly all
state-of-the-art speech recognition systems. Building neural network acoustic
models requires several design decisions including network architecture, size,
and training loss function. This paper offers an empirical investigation on
which aspects of DNN acoustic model design are most important for speech
recognition system performance. We report DNN classifier performance and final
speech recognizer word error rates, and compare DNNs using several metrics to
quantify factors influencing differences in task performance. Our first set of
experiments use the standard Switchboard benchmark corpus, which contains
approximately 300 hours of conversational telephone speech. We compare standard
DNNs to convolutional networks, and present the first experiments using
locally-connected, untied neural networks for acoustic modeling. We
additionally build systems on a corpus of 2,100 hours of training data by
combining the Switchboard and Fisher corpora. This larger corpus allows us to
more thoroughly examine performance of large DNN models -- with up to ten times
more parameters than those typically used in speech recognition systems. Our
results suggest that a relatively simple DNN architecture and optimization
technique produces strong results. These findings, along with previous work,
help establish a set of best practices for building DNN hybrid speech
recognition systems with maximum likelihood training. Our experiments in DNN
optimization additionally serve as a case study for training DNNs with
discriminative loss functions for speech tasks, as well as DNN classifiers more
generally.
|
Learning Laplacian Matrix in Smooth Graph Signal Representations | The construction of a meaningful graph plays a crucial role in the success of
many graph-based representations and algorithms for handling structured data,
especially in the emerging field of graph signal processing. However, a
meaningful graph is not always readily available from the data, nor easy to
define depending on the application domain. In particular, it is often
desirable in graph signal processing applications that a graph is chosen such
that the data admit certain regularity or smoothness on the graph. In this
paper, we address the problem of learning graph Laplacians, which is equivalent
to learning graph topologies, such that the input data form graph signals with
smooth variations on the resulting topology. To this end, we adopt a factor
analysis model for the graph signals and impose a Gaussian probabilistic prior
on the latent variables that control these signals. We show that the Gaussian
prior leads to an efficient representation that favors the smoothness property
of the graph signals. We then propose an algorithm for learning graphs that
enforces such property and is based on minimizing the variations of the signals
on the learned graph. Experiments on both synthetic and real world data
demonstrate that the proposed graph learning framework can efficiently infer
meaningful graph topologies from signal observations under the smoothness
prior.
|
Simple connectome inference from partial correlation statistics in
calcium imaging | In this work, we propose a simple yet effective solution to the problem of
connectome inference in calcium imaging data. The proposed algorithm consists
of two steps. First, processing the raw signals to detect neural peak
activities. Second, inferring the degree of association between neurons from
partial correlation statistics. This paper summarises the methodology that led
us to win the Connectomics Challenge, proposes a simplified version of our
method, and finally compares our results with respect to other inference
methods.
|
Relevance Singular Vector Machine for low-rank matrix sensing | In this paper we develop a new Bayesian inference method for low rank matrix
reconstruction. We call the new method the Relevance Singular Vector Machine
(RSVM) where appropriate priors are defined on the singular vectors of the
underlying matrix to promote low rank. To accelerate computations, a
numerically efficient approximation is developed. The proposed algorithms are
applied to matrix completion and matrix reconstruction problems and their
performance is studied numerically.
|
Rates of Convergence for Nearest Neighbor Classification | Nearest neighbor methods are a popular class of nonparametric estimators with
several desirable properties, such as adaptivity to different distance scales
in different regions of space. Prior work on convergence rates for nearest
neighbor classification has not fully reflected these subtle properties. We
analyze the behavior of these estimators in metric spaces and provide
finite-sample, distribution-dependent rates of convergence under minimal
assumptions. As a by-product, we are able to establish the universal
consistency of nearest neighbor in a broader range of data spaces than was
previously known. We illustrate our upper and lower bounds by introducing
smoothness classes that are customized for nearest neighbor classification.
|
Randomized Block Coordinate Descent for Online and Stochastic
Optimization | Two types of low cost-per-iteration gradient descent methods have been
extensively studied in parallel. One is online or stochastic gradient descent
(OGD/SGD), and the other is randomzied coordinate descent (RBCD). In this
paper, we combine the two types of methods together and propose online
randomized block coordinate descent (ORBCD). At each iteration, ORBCD only
computes the partial gradient of one block coordinate of one mini-batch
samples. ORBCD is well suited for the composite minimization problem where one
function is the average of the losses of a large number of samples and the
other is a simple regularizer defined on high dimensional variables. We show
that the iteration complexity of ORBCD has the same order as OGD or SGD. For
strongly convex functions, by reducing the variance of stochastic gradients, we
show that ORBCD can converge at a geometric rate in expectation, matching the
convergence rate of SGD with variance reduction and RBCD.
|
Mind the Nuisance: Gaussian Process Classification using Privileged
Noise | The learning with privileged information setting has recently attracted a lot
of attention within the machine learning community, as it allows the
integration of additional knowledge into the training process of a classifier,
even when this comes in the form of a data modality that is not available at
test time. Here, we show that privileged information can naturally be treated
as noise in the latent function of a Gaussian Process classifier (GPC). That
is, in contrast to the standard GPC setting, the latent function is not just a
nuisance but a feature: it becomes a natural measure of confidence about the
training data by modulating the slope of the GPC sigmoid likelihood function.
Extensive experiments on public datasets show that the proposed GPC method
using privileged noise, called GPC+, improves over a standard GPC without
privileged knowledge, and also over the current state-of-the-art SVM-based
method, SVM+. Moreover, we show that advanced neural networks and deep learning
methods can be compressed as privileged information.
|
SAGA: A Fast Incremental Gradient Method With Support for Non-Strongly
Convex Composite Objectives | In this work we introduce a new optimisation method called SAGA in the spirit
of SAG, SDCA, MISO and SVRG, a set of recently proposed incremental gradient
algorithms with fast linear convergence rates. SAGA improves on the theory
behind SAG and SVRG, with better theoretical convergence rates, and has support
for composite objectives where a proximal operator is used on the regulariser.
Unlike SDCA, SAGA supports non-strongly convex problems directly, and is
adaptive to any inherent strong convexity of the problem. We give experimental
results showing the effectiveness of our method.
|
A Bayes consistent 1-NN classifier | We show that a simple modification of the 1-nearest neighbor classifier
yields a strongly Bayes consistent learner. Prior to this work, the only
strongly Bayes consistent proximity-based method was the k-nearest neighbor
classifier, for k growing appropriately with sample size. We will argue that a
margin-regularized 1-NN enjoys considerable statistical and algorithmic
advantages over the k-NN classifier. These include user-friendly finite-sample
error bounds, as well as time- and memory-efficient learning and test-point
evaluation algorithms with a principled speed-accuracy tradeoff. Encouraging
empirical results are reported.
|
DC approximation approaches for sparse optimization | Sparse optimization refers to an optimization problem involving the zero-norm
in objective or constraints. In this paper, nonconvex approximation approaches
for sparse optimization have been studied with a unifying point of view in DC
(Difference of Convex functions) programming framework. Considering a common DC
approximation of the zero-norm including all standard sparse inducing penalty
functions, we studied the consistency between global minimums (resp. local
minimums) of approximate and original problems. We showed that, in several
cases, some global minimizers (resp. local minimizers) of the approximate
problem are also those of the original problem. Using exact penalty techniques
in DC programming, we proved stronger results for some particular
approximations, namely, the approximate problem, with suitable parameters, is
equivalent to the original problem. The efficiency of several sparse inducing
penalty functions have been fully analyzed. Four DCA (DC Algorithm) schemes
were developed that cover all standard algorithms in nonconvex sparse
approximation approaches as special versions. They can be viewed as, an $\ell
_{1}$-perturbed algorithm / reweighted-$\ell _{1}$ algorithm / reweighted-$\ell
_{1}$ algorithm. We offer a unifying nonconvex approximation approach, with
solid theoretical tools as well as efficient algorithms based on DC programming
and DCA, to tackle the zero-norm and sparse optimization. As an application, we
implemented our methods for the feature selection in SVM (Support Vector
Machine) problem and performed empirical comparative numerical experiments on
the proposed algorithms with various approximation functions.
|
Identifying Outliers in Large Matrices via Randomized Adaptive
Compressive Sampling | This paper examines the problem of locating outlier columns in a large,
otherwise low-rank, matrix. We propose a simple two-step adaptive sensing and
inference approach and establish theoretical guarantees for its performance;
our results show that accurate outlier identification is achievable using very
few linear summaries of the original data matrix -- as few as the squared rank
of the low-rank component plus the number of outliers, times constant and
logarithmic factors. We demonstrate the performance of our approach
experimentally in two stylized applications, one motivated by robust
collaborative filtering tasks, and the other by saliency map estimation tasks
arising in computer vision and automated surveillance, and also investigate
extensions to settings where the data are noisy, or possibly incomplete.
|
Significant Subgraph Mining with Multiple Testing Correction | The problem of finding itemsets that are statistically significantly enriched
in a class of transactions is complicated by the need to correct for multiple
hypothesis testing. Pruning untestable hypotheses was recently proposed as a
strategy for this task of significant itemset mining. It was shown to lead to
greater statistical power, the discovery of more truly significant itemsets,
than the standard Bonferroni correction on real-world datasets. An open
question, however, is whether this strategy of excluding untestable hypotheses
also leads to greater statistical power in subgraph mining, in which the number
of hypotheses is much larger than in itemset mining. Here we answer this
question by an empirical investigation on eight popular graph benchmark
datasets. We propose a new efficient search strategy, which always returns the
same solution as the state-of-the-art approach and is approximately two orders
of magnitude faster. Moreover, we exploit the dependence between subgraphs by
considering the effective number of tests and thereby further increase the
statistical power.
|
A Multi Level Data Fusion Approach for Speaker Identification on
Telephone Speech | Several speaker identification systems are giving good performance with clean
speech but are affected by the degradations introduced by noisy audio
conditions. To deal with this problem, we investigate the use of complementary
information at different levels for computing a combined match score for the
unknown speaker. In this work, we observe the effect of two supervised machine
learning approaches including support vectors machines (SVM) and na\"ive bayes
(NB). We define two feature vector sets based on mel frequency cepstral
coefficients (MFCC) and relative spectral perceptual linear predictive
coefficients (RASTA-PLP). Each feature is modeled using the Gaussian Mixture
Model (GMM). Several ways of combining these information sources give
significant improvements in a text-independent speaker identification task
using a very large telephone degraded NTIMIT database.
|
Geometric Tight Frame based Stylometry for Art Authentication of van
Gogh Paintings | This paper is about authenticating genuine van Gogh paintings from forgeries.
The authentication process depends on two key steps: feature extraction and
outlier detection. In this paper, a geometric tight frame and some simple
statistics of the tight frame coefficients are used to extract features from
the paintings. Then a forward stage-wise rank boosting is used to select a
small set of features for more accurate classification so that van Gogh
paintings are highly concentrated towards some center point while forgeries are
spread out as outliers. Numerical results show that our method can achieve
86.08% classification accuracy under the leave-one-out cross-validation
procedure. Our method also identifies five features that are much more
predominant than other features. Using just these five features for
classification, our method can give 88.61% classification accuracy which is the
highest so far reported in literature. Evaluation of the five features is also
performed on two hundred datasets generated by bootstrap sampling with
replacement. The median and the mean are 88.61% and 87.77% respectively. Our
results show that a small set of statistics of the tight frame coefficients
along certain orientations can serve as discriminative features for van Gogh
paintings. It is more important to look at the tail distributions of such
directional coefficients than mean values and standard deviations. It reflects
a highly consistent style in van Gogh's brushstroke movements, where many
forgeries demonstrate a more diverse spread in these features.
|
Classification-based Approximate Policy Iteration: Experiments and
Extended Discussions | Tackling large approximate dynamic programming or reinforcement learning
problems requires methods that can exploit regularities, or intrinsic
structure, of the problem in hand. Most current methods are geared towards
exploiting the regularities of either the value function or the policy. We
introduce a general classification-based approximate policy iteration (CAPI)
framework, which encompasses a large class of algorithms that can exploit
regularities of both the value function and the policy space, depending on what
is advantageous. This framework has two main components: a generic value
function estimator and a classifier that learns a policy based on the estimated
value function. We establish theoretical guarantees for the sample complexity
of CAPI-style algorithms, which allow the policy evaluation step to be
performed by a wide variety of algorithms (including temporal-difference-style
methods), and can handle nonparametric representations of policies. Our bounds
on the estimation error of the performance loss are tighter than existing
results. We also illustrate this approach empirically on several problems,
including a large HIV control task.
|
How Many Dissimilarity/Kernel Self Organizing Map Variants Do We Need? | In numerous applicative contexts, data are too rich and too complex to be
represented by numerical vectors. A general approach to extend machine learning
and data mining techniques to such data is to really on a dissimilarity or on a
kernel that measures how different or similar two objects are. This approach
has been used to define several variants of the Self Organizing Map (SOM). This
paper reviews those variants in using a common set of notations in order to
outline differences and similarities between them. It discusses the advantages
and drawbacks of the variants, as well as the actual relevance of the
dissimilarity/kernel SOM for practical applications.
|
Nonparametric Hierarchical Clustering of Functional Data | In this paper, we deal with the problem of curves clustering. We propose a
nonparametric method which partitions the curves into clusters and discretizes
the dimensions of the curve points into intervals. The cross-product of these
partitions forms a data-grid which is obtained using a Bayesian model selection
approach while making no assumptions regarding the curves. Finally, a
post-processing technique, aiming at reducing the number of clusters in order
to improve the interpretability of the clustering, is proposed. It consists in
optimally merging the clusters step by step, which corresponds to an
agglomerative hierarchical classification whose dissimilarity measure is the
variation of the criterion. Interestingly this measure is none other than the
sum of the Kullback-Leibler divergences between clusters distributions before
and after the merges. The practical interest of the approach for functional
data exploratory analysis is presented and compared with an alternative
approach on an artificial and a real world data set.
|
Fast Algorithm for Low-rank matrix recovery in Poisson noise | This paper describes a fast algorithm for recovering low-rank matrices from
their linear measurements contaminated with Poisson noise: the Poisson noise
Maximum Likelihood Singular Value thresholding (PMLSV) algorithm. We propose a
convex optimization formulation with a cost function consisting of the sum of a
likelihood function and a regularization function which the nuclear norm of the
matrix. Instead of solving the optimization problem directly by semi-definite
program (SDP), we derive an iterative singular value thresholding algorithm by
expanding the likelihood function. We demonstrate the good performance of the
proposed algorithm on recovery of solar flare images with Poisson noise: the
algorithm is more efficient than solving SDP using the interior-point algorithm
and it generates a good approximate solution compared to that solved from SDP.
|
Projecting Ising Model Parameters for Fast Mixing | Inference in general Ising models is difficult, due to high treewidth making
tree-based algorithms intractable. Moreover, when interactions are strong,
Gibbs sampling may take exponential time to converge to the stationary
distribution. We present an algorithm to project Ising model parameters onto a
parameter set that is guaranteed to be fast mixing, under several divergences.
We find that Gibbs sampling using the projected parameters is more accurate
than with the original parameters when interaction strengths are strong and
when limited time is available for sampling.
|
Global convergence of splitting methods for nonconvex composite
optimization | We consider the problem of minimizing the sum of a smooth function $h$ with a
bounded Hessian, and a nonsmooth function. We assume that the latter function
is a composition of a proper closed function $P$ and a surjective linear map
$\cal M$, with the proximal mappings of $\tau P$, $\tau > 0$, simple to
compute. This problem is nonconvex in general and encompasses many important
applications in engineering and machine learning. In this paper, we examined
two types of splitting methods for solving this nonconvex optimization problem:
alternating direction method of multipliers and proximal gradient algorithm.
For the direct adaptation of the alternating direction method of multipliers,
we show that, if the penalty parameter is chosen sufficiently large and the
sequence generated has a cluster point, then it gives a stationary point of the
nonconvex problem. We also establish convergence of the whole sequence under an
additional assumption that the functions $h$ and $P$ are semi-algebraic.
Furthermore, we give simple sufficient conditions to guarantee boundedness of
the sequence generated. These conditions can be satisfied for a wide range of
applications including the least squares problem with the $\ell_{1/2}$
regularization. Finally, when $\cal M$ is the identity so that the proximal
gradient algorithm can be efficiently applied, we show that any cluster point
is stationary under a slightly more flexible constant step-size rule than what
is known in the literature for a nonconvex $h$.
|
Structured Learning via Logistic Regression | A successful approach to structured learning is to write the learning
objective as a joint function of linear parameters and inference messages, and
iterate between updates to each. This paper observes that if the inference
problem is "smoothed" through the addition of entropy terms, for fixed
messages, the learning objective reduces to a traditional (non-structured)
logistic regression problem with respect to parameters. In these logistic
regression problems, each training example has a bias term determined by the
current set of messages. Based on this insight, the structured energy function
can be extended from linear factors to any function class where an "oracle"
exists to minimize a logistic loss.
|
Reducing Offline Evaluation Bias in Recommendation Systems | Recommendation systems have been integrated into the majority of large online
systems. They tailor those systems to individual users by filtering and ranking
information according to user profiles. This adaptation process influences the
way users interact with the system and, as a consequence, increases the
difficulty of evaluating a recommendation algorithm with historical data (via
offline evaluation). This paper analyses this evaluation bias and proposes a
simple item weighting solution that reduces its impact. The efficiency of the
proposed solution is evaluated on real world data extracted from Viadeo
professional social network.
|
Anomaly Detection Based on Aggregation of Indicators | Automatic anomaly detection is a major issue in various areas. Beyond mere
detection, the identification of the origin of the problem that produced the
anomaly is also essential. This paper introduces a general methodology that can
assist human operators who aim at classifying monitoring signals. The main idea
is to leverage expert knowledge by generating a very large number of
indicators. A feature selection method is used to keep only the most
discriminant indicators which are used as inputs of a Naive Bayes classifier.
The parameters of the classifier have been optimized indirectly by the
selection process. Simulated data designed to reproduce some of the anomaly
types observed in real world engines.
|
Online Submodular Maximization under a Matroid Constraint with
Application to Learning Assignments | Which ads should we display in sponsored search in order to maximize our
revenue? How should we dynamically rank information sources to maximize the
value of the ranking? These applications exhibit strong diminishing returns:
Redundancy decreases the marginal utility of each ad or information source. We
show that these and other problems can be formalized as repeatedly selecting an
assignment of items to positions to maximize a sequence of monotone submodular
functions that arrive one by one. We present an efficient algorithm for this
general problem and analyze it in the no-regret model. Our algorithm possesses
strong theoretical guarantees, such as a performance ratio that converges to
the optimal constant of 1 - 1/e. We empirically evaluate our algorithm on two
real-world online optimization problems on the web: ad allocation with
submodular utilities, and dynamically ranking blogs to detect information
cascades. Finally, we present a second algorithm that handles the more general
case in which the feasible sets are given by a matroid constraint, while still
maintaining a 1 - 1/e asymptotic performance ratio.
|
Robust Optimization using Machine Learning for Uncertainty Sets | Our goal is to build robust optimization problems for making decisions based
on complex data from the past. In robust optimization (RO) generally, the goal
is to create a policy for decision-making that is robust to our uncertainty
about the future. In particular, we want our policy to best handle the the
worst possible situation that could arise, out of an uncertainty set of
possible situations. Classically, the uncertainty set is simply chosen by the
user, or it might be estimated in overly simplistic ways with strong
assumptions; whereas in this work, we learn the uncertainty set from data
collected in the past. The past data are drawn randomly from an (unknown)
possibly complicated high-dimensional distribution. We propose a new
uncertainty set design and show how tools from statistical learning theory can
be employed to provide probabilistic guarantees on the robustness of the
policy.
|
Expanding the Family of Grassmannian Kernels: An Embedding Perspective | Modeling videos and image-sets as linear subspaces has proven beneficial for
many visual recognition tasks. However, it also incurs challenges arising from
the fact that linear subspaces do not obey Euclidean geometry, but lie on a
special type of Riemannian manifolds known as Grassmannian. To leverage the
techniques developed for Euclidean spaces (e.g, support vector machines) with
subspaces, several recent studies have proposed to embed the Grassmannian into
a Hilbert space by making use of a positive definite kernel. Unfortunately,
only two Grassmannian kernels are known, none of which -as we will show- is
universal, which limits their ability to approximate a target function
arbitrarily well. Here, we introduce several positive definite Grassmannian
kernels, including universal ones, and demonstrate their superiority over
previously-known kernels in various tasks, such as classification, clustering,
sparse coding and hashing.
|
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.