title
stringlengths 7
246
| abstract
stringlengths 6
3.31k
|
|---|---|
Conditions for Convergence in Regularized Machine Learning Objectives | Analysis of the convergence rates of modern convex optimization algorithms
can be achived through binary means: analysis of emperical convergence, or
analysis of theoretical convergence. These two pathways of capturing
information diverge in efficacy when moving to the world of distributed
computing, due to the introduction of non-intuitive, non-linear slowdowns
associated with broadcasting, and in some cases, gathering operations. Despite
these nuances in the rates of convergence, we can still show the existence of
convergence, and lower bounds for the rates. This paper will serve as a helpful
cheat-sheet for machine learning practitioners encountering this problem class
in the field.
|
Machine learning on images using a string-distance | We present a new method for image feature-extraction which is based on
representing an image by a finite-dimensional vector of distances that measure
how different the image is from a set of image prototypes. We use the recently
introduced Universal Image Distance (UID) \cite{RatsabyChesterIEEE2012} to
compare the similarity between an image and a prototype image. The advantage in
using the UID is the fact that no domain knowledge nor any image analysis need
to be done. Each image is represented by a finite dimensional feature vector
whose components are the UID values between the image and a finite set of image
prototypes from each of the feature categories. The method is automatic since
once the user selects the prototype images, the feature vectors are
automatically calculated without the need to do any image analysis. The
prototype images can be of different size, in particular, different than the
image size. Based on a collection of such cases any supervised or unsupervised
learning algorithm can be used to train and produce an image classifier or
image cluster analysis. In this paper we present the image feature-extraction
method and use it on several supervised and unsupervised learning experiments
for satellite image data.
|
Horizon-Independent Optimal Prediction with Log-Loss in Exponential
Families | We study online learning under logarithmic loss with regular parametric
models. Hedayati and Bartlett (2012b) showed that a Bayesian prediction
strategy with Jeffreys prior and sequential normalized maximum likelihood
(SNML) coincide and are optimal if and only if the latter is exchangeable, and
if and only if the optimal strategy can be calculated without knowing the time
horizon in advance. They put forward the question what families have
exchangeable SNML strategies. This paper fully answers this open problem for
one-dimensional exponential families. The exchangeability can happen only for
three classes of natural exponential family distributions, namely the Gaussian,
Gamma, and the Tweedie exponential family of order 3/2. Keywords: SNML
Exchangeability, Exponential Family, Online Learning, Logarithmic Loss,
Bayesian Strategy, Jeffreys Prior, Fisher Information1
|
Generalized Centroid Estimators in Bioinformatics | In a number of estimation problems in bioinformatics, accuracy measures of
the target problem are usually given, and it is important to design estimators
that are suitable to those accuracy measures. However, there is often a
discrepancy between an employed estimator and a given accuracy measure of the
problem. In this study, we introduce a general class of efficient estimators
for estimation problems on high-dimensional binary spaces, which representmany
fundamental problems in bioinformatics. Theoretical analysis reveals that the
proposed estimators generally fit with commonly-used accuracy measures (e.g.
sensitivity, PPV, MCC and F-score) as well as it can be computed efficiently in
many cases, and cover a wide range of problems in bioinformatics from the
viewpoint of the principle of maximum expected accuracy (MEA). It is also shown
that some important algorithms in bioinformatics can be interpreted in a
unified manner. Not only the concept presented in this paper gives a useful
framework to design MEA-based estimators but also it is highly extendable and
sheds new light on many problems in bioinformatics.
|
Ensembles of Classifiers based on Dimensionality Reduction | We present a novel approach for the construction of ensemble classifiers
based on dimensionality reduction. Dimensionality reduction methods represent
datasets using a small number of attributes while preserving the information
conveyed by the original dataset. The ensemble members are trained based on
dimension-reduced versions of the training set. These versions are obtained by
applying dimensionality reduction to the original training set using different
values of the input parameters. This construction meets both the diversity and
accuracy criteria which are required to construct an ensemble classifier where
the former criterion is obtained by the various input parameter values and the
latter is achieved due to the decorrelation and noise reduction properties of
dimensionality reduction. In order to classify a test sample, it is first
embedded into the dimension reduced space of each individual classifier by
using an out-of-sample extension algorithm. Each classifier is then applied to
the embedded sample and the classification is obtained via a voting scheme. We
present three variations of the proposed approach based on the Random
Projections, the Diffusion Maps and the Random Subspaces dimensionality
reduction algorithms. We also present a multi-strategy ensemble which combines
AdaBoost and Diffusion Maps. A comparison is made with the Bagging, AdaBoost,
Rotation Forest ensemble classifiers and also with the base classifier which
does not incorporate dimensionality reduction. Our experiments used seventeen
benchmark datasets from the UCI repository. The results obtained by the
proposed algorithms were superior in many cases to other algorithms.
|
Meta Path-Based Collective Classification in Heterogeneous Information
Networks | Collective classification has been intensively studied due to its impact in
many important applications, such as web mining, bioinformatics and citation
analysis. Collective classification approaches exploit the dependencies of a
group of linked objects whose class labels are correlated and need to be
predicted simultaneously. In this paper, we focus on studying the collective
classification problem in heterogeneous networks, which involves multiple types
of data objects interconnected by multiple types of links. Intuitively, two
objects are correlated if they are linked by many paths in the network.
However, most existing approaches measure the dependencies among objects
through directly links or indirect links without considering the different
semantic meanings behind different paths. In this paper, we study the
collective classification problem taht is defined among the same type of
objects in heterogenous networks. Moreover, by considering different linkage
paths in the network, one can capture the subtlety of different types of
dependencies among objects. We introduce the concept of meta-path based
dependencies among objects, where a meta path is a path consisting a certain
sequence of linke types. We show that the quality of collective classification
results strongly depends upon the meta paths used. To accommodate the large
network size, a novel solution, called HCC (meta-path based Heterogenous
Collective Classification), is developed to effectively assign labels to a
group of instances that are interconnected through different meta-paths. The
proposed HCC model can capture different types of dependencies among objects
with respect to different meta paths. Empirical studies on real-world networks
demonstrate that effectiveness of the proposed meta path-based collective
classification approach.
|
Robustness of Random Forest-based gene selection methods | Gene selection is an important part of microarray data analysis because it
provides information that can lead to a better mechanistic understanding of an
investigated phenomenon. At the same time, gene selection is very difficult
because of the noisy nature of microarray data. As a consequence, gene
selection is often performed with machine learning methods. The Random Forest
method is particularly well suited for this purpose. In this work, four
state-of-the-art Random Forest-based feature selection methods were compared in
a gene selection context. The analysis focused on the stability of selection
because, although it is necessary for determining the significance of results,
it is often ignored in similar studies.
The comparison of post-selection accuracy in the validation of Random Forest
classifiers revealed that all investigated methods were equivalent in this
context. However, the methods substantially differed with respect to the number
of selected genes and the stability of selection. Of the analysed methods, the
Boruta algorithm predicted the most genes as potentially important.
The post-selection classifier error rate, which is a frequently used measure,
was found to be a potentially deceptive measure of gene selection quality. When
the number of consistently selected genes was considered, the Boruta algorithm
was clearly the best. Although it was also the most computationally intensive
method, the Boruta algorithm's computational demands could be reduced to levels
comparable to those of other algorithms by replacing the Random Forest
importance with a comparable measure from Random Ferns (a similar but
simplified classifier). Despite their design assumptions, the minimal optimal
selection methods, were found to select a high fraction of false positives.
|
On the Complexity Analysis of Randomized Block-Coordinate Descent
Methods | In this paper we analyze the randomized block-coordinate descent (RBCD)
methods proposed in [8,11] for minimizing the sum of a smooth convex function
and a block-separable convex function. In particular, we extend Nesterov's
technique developed in [8] for analyzing the RBCD method for minimizing a
smooth convex function over a block-separable closed convex set to the
aforementioned more general problem and obtain a sharper expected-value type of
convergence rate than the one implied in [11]. Also, we obtain a better
high-probability type of iteration complexity, which improves upon the one in
[11] by at least the amount $O(n/\epsilon)$, where $\epsilon$ is the target
solution accuracy and $n$ is the number of problem blocks. In addition, for
unconstrained smooth convex minimization, we develop a new technique called
{\it randomized estimate sequence} to analyze the accelerated RBCD method
proposed by Nesterov [11] and establish a sharper expected-value type of
convergence rate than the one given in [11].
|
Power to the Points: Validating Data Memberships in Clusterings | A clustering is an implicit assignment of labels of points, based on
proximity to other points. It is these labels that are then used for downstream
analysis (either focusing on individual clusters, or identifying
representatives of clusters and so on). Thus, in order to trust a clustering as
a first step in exploratory data analysis, we must trust the labels assigned to
individual data. Without supervision, how can we validate this assignment? In
this paper, we present a method to attach affinity scores to the implicit
labels of individual points in a clustering. The affinity scores capture the
confidence level of the cluster that claims to "own" the point. This method is
very general: it can be used with clusterings derived from Euclidean data,
kernelized data, or even data derived from information spaces. It smoothly
incorporates importance functions on clusters, allowing us to eight different
clusters differently. It is also efficient: assigning an affinity score to a
point depends only polynomially on the number of clusters and is independent of
the number of points in the data. The dimensionality of the underlying space
only appears in preprocessing. We demonstrate the value of our approach with an
experimental study that illustrates the use of these scores in different data
analysis tasks, as well as the efficiency and flexibility of the method. We
also demonstrate useful visualizations of these scores; these might prove
useful within an interactive analytics framework.
|
Zero-sum repeated games: Counterexamples to the existence of the
asymptotic value and the conjecture
$\operatorname{maxmin}=\operatorname{lim}v_n$ | Mertens [In Proceedings of the International Congress of Mathematicians
(Berkeley, Calif., 1986) (1987) 1528-1577 Amer. Math. Soc.] proposed two
general conjectures about repeated games: the first one is that, in any
two-person zero-sum repeated game, the asymptotic value exists, and the second
one is that, when Player 1 is more informed than Player 2, in the long run
Player 1 is able to guarantee the asymptotic value. We disprove these two
long-standing conjectures by providing an example of a zero-sum repeated game
with public signals and perfect observation of the actions, where the value of
the $\lambda$-discounted game does not converge when $\lambda$ goes to 0. The
aforementioned example involves seven states, two actions and two signals for
each player. Remarkably, players observe the payoffs, and play in turn.
|
A Data Mining Approach to Solve the Goal Scoring Problem | In soccer, scoring goals is a fundamental objective which depends on many
conditions and constraints. Considering the RoboCup soccer 2D-simulator, this
paper presents a data mining-based decision system to identify the best time
and direction to kick the ball towards the goal to maximize the overall chances
of scoring during a simulated soccer match. Following the CRISP-DM methodology,
data for modeling were extracted from matches of major international
tournaments (10691 kicks), knowledge about soccer was embedded via
transformation of variables and a Multilayer Perceptron was used to estimate
the scoring chance. Experimental performance assessment to compare this
approach against previous LDA-based approach was conducted from 100 matches.
Several statistical metrics were used to analyze the performance of the system
and the results showed an increase of 7.7% in the number of kicks, producing an
overall increase of 78% in the number of goals scored.
|
Robust Logistic Regression using Shift Parameters (Long Version) | Annotation errors can significantly hurt classifier performance, yet datasets
are only growing noisier with the increased use of Amazon Mechanical Turk and
techniques like distant supervision that automatically generate labels. In this
paper, we present a robust extension of logistic regression that incorporates
the possibility of mislabelling directly into the objective. Our model can be
trained through nearly the same means as logistic regression, and retains its
efficiency on high-dimensional datasets. Through named entity recognition
experiments, we demonstrate that our approach can provide a significant
improvement over the standard model when annotation errors are present.
|
Divide and Conquer Kernel Ridge Regression: A Distributed Algorithm with
Minimax Optimal Rates | We establish optimal convergence rates for a decomposition-based scalable
approach to kernel ridge regression. The method is simple to describe: it
randomly partitions a dataset of size N into m subsets of equal size, computes
an independent kernel ridge regression estimator for each subset, then averages
the local solutions into a global predictor. This partitioning leads to a
substantial reduction in computation time versus the standard approach of
performing kernel ridge regression on all N samples. Our two main theorems
establish that despite the computational speed-up, statistical optimality is
retained: as long as m is not too large, the partition-based estimator achieves
the statistical minimax rate over all estimators using the set of N samples. As
concrete examples, our theory guarantees that the number of processors m may
grow nearly linearly for finite-rank kernels and Gaussian kernels and
polynomially in N for Sobolev spaces, which in turn allows for substantial
reductions in computational cost. We conclude with experiments on both
simulated data and a music-prediction task that complement our theoretical
results, exhibiting the computational and statistical benefits of our approach.
|
A Comparison of Random Forests and Ferns on Recognition of Instruments
in Jazz Recordings | In this paper, we first apply random ferns for classification of real music
recordings of a jazz band. No initial segmentation of audio data is assumed,
i.e., no onset, offset, nor pitch data are needed. The notion of random ferns
is described in the paper, to familiarize the reader with this classification
algorithm, which was introduced quite recently and applied so far in image
recognition tasks. The performance of random ferns is compared with random
forests for the same data. The results of experiments are presented in the
paper, and conclusions are drawn.
|
A Supervised Neural Autoregressive Topic Model for Simultaneous Image
Classification and Annotation | Topic modeling based on latent Dirichlet allocation (LDA) has been a
framework of choice to perform scene recognition and annotation. Recently, a
new type of topic model called the Document Neural Autoregressive Distribution
Estimator (DocNADE) was proposed and demonstrated state-of-the-art performance
for document modeling. In this work, we show how to successfully apply and
extend this model to the context of visual scene modeling. Specifically, we
propose SupDocNADE, a supervised extension of DocNADE, that increases the
discriminative power of the hidden topic features by incorporating label
information into the training objective of the model. We also describe how to
leverage information about the spatial position of the visual words and how to
embed additional image annotations, so as to simultaneously perform image
classification and annotation. We test our model on the Scene15, LabelMe and
UIUC-Sports datasets and show that it compares favorably to other topic models
such as the supervised variant of LDA.
|
A Primal Condition for Approachability with Partial Monitoring | In approachability with full monitoring there are two types of conditions
that are known to be equivalent for convex sets: a primal and a dual condition.
The primal one is of the form: a set C is approachable if and only all
containing half-spaces are approachable in the one-shot game; while the dual
one is of the form: a convex set C is approachable if and only if it intersects
all payoff sets of a certain form. We consider approachability in games with
partial monitoring. In previous works (Perchet 2011; Mannor et al. 2011) we
provided a dual characterization of approachable convex sets; we also exhibited
efficient strategies in the case where C is a polytope. In this paper we
provide primal conditions on a convex set to be approachable with partial
monitoring. They depend on a modified reward function and lead to
approachability strategies, based on modified payoff functions, that proceed by
projections similarly to Blackwell's (1956) strategy; this is in contrast with
previously studied strategies in this context that relied mostly on the
signaling structure and aimed at estimating well the distributions of the
signals received. Our results generalize classical results by Kohlberg 1975
(see also Mertens et al. 1994) and apply to games with arbitrary signaling
structure as well as to arbitrary convex sets.
|
Characterizing A Database of Sequential Behaviors with Latent Dirichlet
Hidden Markov Models | This paper proposes a generative model, the latent Dirichlet hidden Markov
models (LDHMM), for characterizing a database of sequential behaviors
(sequences). LDHMMs posit that each sequence is generated by an underlying
Markov chain process, which are controlled by the corresponding parameters
(i.e., the initial state vector, transition matrix and the emission matrix).
These sequence-level latent parameters for each sequence are modeled as latent
Dirichlet random variables and parameterized by a set of deterministic
database-level hyper-parameters. Through this way, we expect to model the
sequence in two levels: the database level by deterministic hyper-parameters
and the sequence-level by latent parameters. To learn the deterministic
hyper-parameters and approximate posteriors of parameters in LDHMMs, we propose
an iterative algorithm under the variational EM framework, which consists of E
and M steps. We examine two different schemes, the fully-factorized and
partially-factorized forms, for the framework, based on different assumptions.
We present empirical results of behavior modeling and sequence classification
on three real-world data sets, and compare them to other related models. The
experimental results prove that the proposed LDHMMs produce better
generalization performance in terms of log-likelihood and deliver competitive
results on the sequence classification problem.
|
Adapting the Stochastic Block Model to Edge-Weighted Networks | We generalize the stochastic block model to the important case in which edges
are annotated with weights drawn from an exponential family distribution. This
generalization introduces several technical difficulties for model estimation,
which we solve using a Bayesian approach. We introduce a variational algorithm
that efficiently approximates the model's posterior distribution for dense
graphs. In specific numerical experiments on edge-weighted networks, this
weighted stochastic block model outperforms the common approach of first
applying a single threshold to all weights and then applying the classic
stochastic block model, which can obscure latent block structure in networks.
This model will enable the recovery of latent structure in a broader range of
network data than was previously possible.
|
Parallel Gaussian Process Regression with Low-Rank Covariance Matrix
Approximations | Gaussian processes (GP) are Bayesian non-parametric models that are widely
used for probabilistic regression. Unfortunately, it cannot scale well with
large data nor perform real-time predictions due to its cubic time cost in the
data size. This paper presents two parallel GP regression methods that exploit
low-rank covariance matrix approximations for distributing the computational
load among parallel machines to achieve time efficiency and scalability. We
theoretically guarantee the predictive performances of our proposed parallel
GPs to be equivalent to that of some centralized approximate GP regression
methods: The computation of their centralized counterparts can be distributed
among parallel machines, hence achieving greater time efficiency and
scalability. We analytically compare the properties of our parallel GPs such as
time, space, and communication complexity. Empirical evaluation on two
real-world datasets in a cluster of 20 computing nodes shows that our parallel
GPs are significantly more time-efficient and scalable than their centralized
counterparts and exact/full GP while achieving predictive performances
comparable to full GP.
|
A Symmetric Rank-one Quasi Newton Method for Non-negative Matrix
Factorization | As we all known, the nonnegative matrix factorization (NMF) is a dimension
reduction method that has been widely used in image processing, text
compressing and signal processing etc. In this paper, an algorithm for
nonnegative matrix approximation is proposed. This method mainly bases on the
active set and the quasi-Newton type algorithm, by using the symmetric rank-one
and negative curvature direction technologies to approximate the Hessian
matrix. Our method improves the recent results of those methods in [Pattern
Recognition, 45(2012)3557-3565; SIAM J. Sci. Comput., 33(6)(2011)3261-3281;
Neural Computation, 19(10)(2007)2756-2779, etc.]. Moreover, the object function
decreases faster than many other NMF methods. In addition, some numerical
experiments are presented in the synthetic data, imaging processing and text
clustering. By comparing with the other six nonnegative matrix approximation
methods, our experiments confirm to our analysis.
|
Supervised Feature Selection for Diagnosis of Coronary Artery Disease
Based on Genetic Algorithm | Feature Selection (FS) has become the focus of much research on decision
support systems areas for which data sets with tremendous number of variables
are analyzed. In this paper we present a new method for the diagnosis of
Coronary Artery Diseases (CAD) founded on Genetic Algorithm (GA) wrapped Bayes
Naive (BN) based FS. Basically, CAD dataset contains two classes defined with
13 features. In GA BN algorithm, GA generates in each iteration a subset of
attributes that will be evaluated using the BN in the second step of the
selection procedure. The final set of attribute contains the most relevant
feature model that increases the accuracy. The algorithm in this case produces
85.50% classification accuracy in the diagnosis of CAD. Thus, the asset of the
Algorithm is then compared with the use of Support Vector Machine (SVM),
MultiLayer Perceptron (MLP) and C4.5 decision tree Algorithm. The result of
classification accuracy for those algorithms are respectively 83.5%, 83.16% and
80.85%. Consequently, the GA wrapped BN Algorithm is correspondingly compared
with other FS algorithms. The Obtained results have shown very promising
outcomes for the diagnosis of CAD.
|
Information-Theoretic Approach to Efficient Adaptive Path Planning for
Mobile Robotic Environmental Sensing | Recent research in robot exploration and mapping has focused on sampling
environmental hotspot fields. This exploration task is formalized by Low,
Dolan, and Khosla (2008) in a sequential decision-theoretic planning under
uncertainty framework called MASP. The time complexity of solving MASP
approximately depends on the map resolution, which limits its use in
large-scale, high-resolution exploration and mapping. To alleviate this
computational difficulty, this paper presents an information-theoretic approach
to MASP (iMASP) for efficient adaptive path planning; by reformulating the
cost-minimizing iMASP as a reward-maximizing problem, its time complexity
becomes independent of map resolution and is less sensitive to increasing robot
team size as demonstrated both theoretically and empirically. Using the
reward-maximizing dual, we derive a novel adaptive variant of maximum entropy
sampling, thus improving the induced exploration policy performance. It also
allows us to establish theoretical bounds quantifying the performance advantage
of optimal adaptive over non-adaptive policies and the performance quality of
approximately optimal vs. optimal adaptive policies. We show analytically and
empirically the superior performance of iMASP-based policies for sampling the
log-Gaussian process to that of policies for the widely-used Gaussian process
in mapping the hotspot field. Lastly, we provide sufficient conditions that,
when met, guarantee adaptivity has no benefit under an assumed environment
model.
|
Fast and accurate sentiment classification using an enhanced Naive Bayes
model | We have explored different methods of improving the accuracy of a Naive Bayes
classifier for sentiment analysis. We observed that a combination of methods
like negation handling, word n-grams and feature selection by mutual
information results in a significant improvement in accuracy. This implies that
a highly accurate and fast sentiment classifier can be built using a simple
Naive Bayes model that has linear training and testing time complexities. We
achieved an accuracy of 88.80% on the popular IMDB movie reviews dataset.
|
Optimal rates of convergence for persistence diagrams in Topological
Data Analysis | Computational topology has recently known an important development toward
data analysis, giving birth to the field of topological data analysis.
Topological persistence, or persistent homology, appears as a fundamental tool
in this field. In this paper, we study topological persistence in general
metric spaces, with a statistical approach. We show that the use of persistent
homology can be naturally considered in general statistical frameworks and
persistence diagrams can be used as statistics with interesting convergence
properties. Some numerical experiments are performed in various contexts to
illustrate our results.
|
Reinforcement Learning for the Soccer Dribbling Task | We propose a reinforcement learning solution to the \emph{soccer dribbling
task}, a scenario in which a soccer agent has to go from the beginning to the
end of a region keeping possession of the ball, as an adversary attempts to
gain possession. While the adversary uses a stationary policy, the dribbler
learns the best action to take at each decision point. After defining
meaningful variables to represent the state space, and high-level macro-actions
to incorporate domain knowledge, we describe our application of the
reinforcement learning algorithm \emph{Sarsa} with CMAC for function
approximation. Our experiments show that, after the training period, the
dribbler is able to accomplish its task against a strong adversary around 58%
of the time.
|
Normalized Online Learning | We introduce online learning algorithms which are independent of feature
scales, proving regret bounds dependent on the ratio of scales existent in the
data rather than the absolute scale. This has several useful effects: there is
no need to pre-normalize data, the test-time and test-space complexity are
reduced, and the algorithms are more robust.
|
Dynamic Clustering via Asymptotics of the Dependent Dirichlet Process
Mixture | This paper presents a novel algorithm, based upon the dependent Dirichlet
process mixture model (DDPMM), for clustering batch-sequential data containing
an unknown number of evolving clusters. The algorithm is derived via a
low-variance asymptotic analysis of the Gibbs sampling algorithm for the DDPMM,
and provides a hard clustering with convergence guarantees similar to those of
the k-means algorithm. Empirical results from a synthetic test with moving
Gaussian clusters and a test with real ADS-B aircraft trajectory data
demonstrate that the algorithm requires orders of magnitude less computational
time than contemporary probabilistic and hard clustering algorithms, while
providing higher accuracy on the examined datasets.
|
Generalized Denoising Auto-Encoders as Generative Models | Recent work has shown how denoising and contractive autoencoders implicitly
capture the structure of the data-generating density, in the case where the
corruption noise is Gaussian, the reconstruction error is the squared error,
and the data is continuous-valued. This has led to various proposals for
sampling from this implicitly learned density function, using Langevin and
Metropolis-Hastings MCMC. However, it remained unclear how to connect the
training procedure of regularized auto-encoders to the implicit estimation of
the underlying data-generating distribution when the data are discrete, or
using other forms of corruption process and reconstruction errors. Another
issue is the mathematical justification which is only valid in the limit of
small corruption noise. We propose here a different attack on the problem,
which deals with all these issues: arbitrary (but noisy enough) corruption,
arbitrary reconstruction loss (seen as a log-likelihood), handling both
discrete and continuous-valued variables, and removing the bias due to
non-infinitesimal corruption noise (or non-infinitesimal contractive penalty).
|
Predicting the Severity of Breast Masses with Data Mining Methods | Mammography is the most effective and available tool for breast cancer
screening. However, the low positive predictive value of breast biopsy
resulting from mammogram interpretation leads to approximately 70% unnecessary
biopsies with benign outcomes. Data mining algorithms could be used to help
physicians in their decisions to perform a breast biopsy on a suspicious lesion
seen in a mammogram image or to perform a short term follow-up examination
instead. In this research paper data mining classification algorithms; Decision
Tree (DT), Artificial Neural Network (ANN), and Support Vector Machine (SVM)
are analyzed on mammographic masses data set. The purpose of this study is to
increase the ability of physicians to determine the severity (benign or
malignant) of a mammographic mass lesion from BI-RADS attributes and the
patient,s age. The whole data set is divided for training the models and test
them by the ratio of 70:30% respectively and the performances of classification
algorithms are compared through three statistical measures; sensitivity,
specificity, and classification accuracy. Accuracy of DT, ANN and SVM are
78.12%, 80.56% and 81.25% of test samples respectively. Our analysis shows that
out of these three classification models SVM predicts severity of breast cancer
with least error rate and highest accuracy.
|
Test cost and misclassification cost trade-off using reframing | Many solutions to cost-sensitive classification (and regression) rely on some
or all of the following assumptions: we have complete knowledge about the cost
context at training time, we can easily re-train whenever the cost context
changes, and we have technique-specific methods (such as cost-sensitive
decision trees) that can take advantage of that information. In this paper we
address the problem of selecting models and minimising joint cost (integrating
both misclassification cost and test costs) without any of the above
assumptions. We introduce methods and plots (such as the so-called JROC plots)
that can work with any off-the-shelf predictive technique, including ensembles,
such that we reframe the model to use the appropriate subset of attributes (the
feature configuration) during deployment time. In other words, models are
trained with the available attributes (once and for all) and then deployed by
setting missing values on the attributes that are deemed ineffective for
reducing the joint cost. As the number of feature configuration combinations
grows exponentially with the number of features we introduce quadratic methods
that are able to approximate the optimal configuration and model choices, as
shown by the experimental results.
|
Alternating Decision trees for early diagnosis of dengue fever | Dengue fever is a flu-like illness spread by the bite of an infected mosquito
which is fast emerging as a major health problem. Timely and cost effective
diagnosis using clinical and laboratory features would reduce the mortality
rates besides providing better grounds for clinical management and disease
surveillance. We wish to develop a robust and effective decision tree based
approach for predicting dengue disease. Our analysis is based on the clinical
characteristics and laboratory measurements of the diseased individuals. We
have developed and trained an alternating decision tree with boosting and
compared its performance with C4.5 algorithm for dengue disease diagnosis. Of
the 65 patient records a diagnosis establishes that 53 individuals have been
confirmed to have dengue fever. An alternating decision tree based algorithm
was able to differentiate the dengue fever using the clinical and laboratory
data with number of correctly classified instances as 89%, F-measure of 0.86
and receiver operator characteristics (ROC) of 0.826 as compared to C4.5 having
correctly classified instances as 78%,h F-measure of 0.738 and ROC of 0.617
respectively. Alternating decision tree based approach with boosting has been
able to predict dengue fever with a higher degree of accuracy than C4.5 based
decision tree using simple clinical and laboratory features. Further analysis
on larger data sets is required to improve the sensitivity and specificity of
the alternating decision trees.
|
Privileged Information for Data Clustering | Many machine learning algorithms assume that all input samples are
independently and identically distributed from some common distribution on
either the input space X, in the case of unsupervised learning, or the input
and output space X x Y in the case of supervised and semi-supervised learning.
In the last number of years the relaxation of this assumption has been explored
and the importance of incorporation of additional information within machine
learning algorithms became more apparent. Traditionally such fusion of
information was the domain of semi-supervised learning. More recently the
inclusion of knowledge from separate hypothetical spaces has been proposed by
Vapnik as part of the supervised setting. In this work we are interested in
exploring Vapnik's idea of master-class learning and the associated learning
using privileged information, however within the unsupervised setting. Adoption
of the advanced supervised learning paradigm for the unsupervised setting
instigates investigation into the difference between privileged and technical
data. By means of our proposed aRi-MAX method stability of the KMeans algorithm
is improved and identification of the best clustering solution is achieved on
an artificial dataset. Subsequently an information theoretic dot product based
algorithm called P-Dot is proposed. This method has the ability to utilize a
wide variety of clustering techniques, individually or in combination, while
fusing privileged and technical data for improved clustering. Application of
the P-Dot method to the task of digit recognition confirms our findings in a
real-world scenario.
|
On model selection consistency of regularized M-estimators | Regularized M-estimators are used in diverse areas of science and engineering
to fit high-dimensional models with some low-dimensional structure. Usually the
low-dimensional structure is encoded by the presence of the (unknown)
parameters in some low-dimensional model subspace. In such settings, it is
desirable for estimates of the model parameters to be \emph{model selection
consistent}: the estimates also fall in the model subspace. We develop a
general framework for establishing consistency and model selection consistency
of regularized M-estimators and show how it applies to some special cases of
interest in statistical learning. Our analysis identifies two key properties of
regularized M-estimators, referred to as geometric decomposability and
irrepresentability, that ensure the estimators are consistent and model
selection consistent.
|
Understanding ACT-R - an Outsider's Perspective | The ACT-R theory of cognition developed by John Anderson and colleagues
endeavors to explain how humans recall chunks of information and how they solve
problems. ACT-R also serves as a theoretical basis for "cognitive tutors",
i.e., automatic tutoring systems that help students learn mathematics, computer
programming, and other subjects. The official ACT-R definition is distributed
across a large body of literature spanning many articles and monographs, and
hence it is difficult for an "outsider" to learn the most important aspects of
the theory. This paper aims to provide a tutorial to the core components of the
ACT-R theory.
|
Dynamic Ad Allocation: Bandits with Budgets | We consider an application of multi-armed bandits to internet advertising
(specifically, to dynamic ad allocation in the pay-per-click model, with
uncertainty on the click probabilities). We focus on an important practical
issue that advertisers are constrained in how much money they can spend on
their ad campaigns. This issue has not been considered in the prior work on
bandit-based approaches for ad allocation, to the best of our knowledge.
We define a simple, stylized model where an algorithm picks one ad to display
in each round, and each ad has a \emph{budget}: the maximal amount of money
that can be spent on this ad. This model admits a natural variant of UCB1, a
well-known algorithm for multi-armed bandits with stochastic rewards. We derive
strong provable guarantees for this algorithm.
|
Phase Retrieval using Alternating Minimization | Phase retrieval problems involve solving linear equations, but with missing
sign (or phase, for complex numbers) information. More than four decades after
it was first proposed, the seminal error reduction algorithm of (Gerchberg and
Saxton 1972) and (Fienup 1982) is still the popular choice for solving many
variants of this problem. The algorithm is based on alternating minimization;
i.e. it alternates between estimating the missing phase information, and the
candidate solution. Despite its wide usage in practice, no global convergence
guarantees for this algorithm are known. In this paper, we show that a
(resampling) variant of this approach converges geometrically to the solution
of one such problem -- finding a vector $\mathbf{x}$ from
$\mathbf{y},\mathbf{A}$, where $\mathbf{y} =
\left|\mathbf{A}^{\top}\mathbf{x}\right|$ and $|\mathbf{z}|$ denotes a vector
of element-wise magnitudes of $\mathbf{z}$ -- under the assumption that
$\mathbf{A}$ is Gaussian.
Empirically, we demonstrate that alternating minimization performs similar to
recently proposed convex techniques for this problem (which are based on
"lifting" to a convex matrix problem) in sample complexity and robustness to
noise. However, it is much more efficient and can scale to large problems.
Analytically, for a resampling version of alternating minimization, we show
geometric convergence to the solution, and sample complexity that is off by log
factors from obvious lower bounds. We also establish close to optimal scaling
for the case when the unknown vector is sparse. Our work represents the first
theoretical guarantee for alternating minimization (albeit with resampling) for
any variant of phase retrieval problems in the non-convex setting.
|
RNADE: The real-valued neural autoregressive density-estimator | We introduce RNADE, a new model for joint density estimation of real-valued
vectors. Our model calculates the density of a datapoint as the product of
one-dimensional conditionals modeled using mixture density networks with shared
parameters. RNADE learns a distributed representation of the data, while having
a tractable expression for the calculation of densities. A tractable likelihood
allows direct comparison with other methods and training by standard
gradient-based optimizers. We compare the performance of RNADE on several
datasets of heterogeneous and perceptual data, finding it outperforms mixture
models in all but one case.
|
Guided Random Forest in the RRF Package | Random Forest (RF) is a powerful supervised learner and has been popularly
used in many applications such as bioinformatics.
In this work we propose the guided random forest (GRF) for feature selection.
Similar to a feature selection method called guided regularized random forest
(GRRF), GRF is built using the importance scores from an ordinary RF. However,
the trees in GRRF are built sequentially, are highly correlated and do not
allow for parallel computing, while the trees in GRF are built independently
and can be implemented in parallel. Experiments on 10 high-dimensional gene
data sets show that, with a fixed parameter value (without tuning the
parameter), RF applied to features selected by GRF outperforms RF applied to
all features on 9 data sets and 7 of them have significant differences at the
0.05 level. Therefore, both accuracy and interpretability are significantly
improved. GRF selects more features than GRRF, however, leads to better
classification accuracy. Note in this work the guided random forest is guided
by the importance scores from an ordinary random forest, however, it can also
be guided by other methods such as human insights (by specifying $\lambda_i$).
GRF can be used in "RRF" v1.4 (and later versions), a package that also
includes the regularized random forest methods.
|
Deep Learning using Linear Support Vector Machines | Recently, fully-connected and convolutional neural networks have been trained
to achieve state-of-the-art performance on a wide variety of tasks such as
speech recognition, image classification, natural language processing, and
bioinformatics. For classification tasks, most of these "deep learning" models
employ the softmax activation function for prediction and minimize
cross-entropy loss. In this paper, we demonstrate a small but consistent
advantage of replacing the softmax layer with a linear support vector machine.
Learning minimizes a margin-based loss instead of the cross-entropy loss. While
there have been various combinations of neural nets and SVMs in prior art, our
results using L2-SVMs show that by simply replacing softmax with linear SVMs
gives significant gains on popular deep learning datasets MNIST, CIFAR-10, and
the ICML 2013 Representation Learning Workshop's face expression recognition
challenge.
|
KERT: Automatic Extraction and Ranking of Topical Keyphrases from
Content-Representative Document Titles | We introduce KERT (Keyphrase Extraction and Ranking by Topic), a framework
for topical keyphrase generation and ranking. By shifting from the
unigram-centric traditional methods of unsupervised keyphrase extraction to a
phrase-centric approach, we are able to directly compare and rank phrases of
different lengths. We construct a topical keyphrase ranking function which
implements the four criteria that represent high quality topical keyphrases
(coverage, purity, phraseness, and completeness). The effectiveness of our
approach is demonstrated on two collections of content-representative titles in
the domains of Computer Science and Physics.
|
Probabilistic Solutions to Differential Equations and their Application
to Riemannian Statistics | We study a probabilistic numerical method for the solution of both boundary
and initial value problems that returns a joint Gaussian process posterior over
the solution. Such methods have concrete value in the statistics on Riemannian
manifolds, where non-analytic ordinary differential equations are involved in
virtually all computations. The probabilistic formulation permits marginalising
the uncertainty of the numerical solution such that statistics are less
sensitive to inaccuracies. This leads to new Riemannian algorithms for mean
value computations and principal geodesic analysis. Marginalisation also means
results can be less precise than point estimates, enabling a noticeable
speed-up over the state of the art. Our approach is an argument for a wider
point that uncertainty caused by numerical calculations should be tracked
throughout the pipeline of machine learning algorithms.
|
Learning from networked examples in a k-partite graph | Many machine learning algorithms are based on the assumption that training
examples are drawn independently. However, this assumption does not hold
anymore when learning from a networked sample where two or more training
examples may share common features. We propose an efficient weighting method
for learning from networked examples and show the sample error bound which is
better than previous work.
|
Riemannian metrics for neural networks II: recurrent networks and
learning symbolic data sequences | Recurrent neural networks are powerful models for sequential data, able to
represent complex dependencies in the sequence that simpler models such as
hidden Markov models cannot handle. Yet they are notoriously hard to train.
Here we introduce a training procedure using a gradient ascent in a Riemannian
metric: this produces an algorithm independent from design choices such as the
encoding of parameters and unit activities. This metric gradient ascent is
designed to have an algorithmic cost close to backpropagation through time for
sparsely connected networks. We use this procedure on gated leaky neural
networks (GLNNs), a variant of recurrent neural networks with an architecture
inspired by finite automata and an evolution equation inspired by
continuous-time networks. GLNNs trained with a Riemannian gradient are
demonstrated to effectively capture a variety of structures in synthetic
problems: basic block nesting as in context-free grammars (an important feature
of natural languages, but difficult to learn), intersections of multiple
independent Markov-type relations, or long-distance relationships such as the
distant-XOR problem. This method does not require adjusting the network
structure or initial parameters: the network used is a sparse random graph and
the initialization is identical for all problems considered.
|
On the Performance Bounds of some Policy Search Dynamic Programming
Algorithms | We consider the infinite-horizon discounted optimal control problem
formalized by Markov Decision Processes. We focus on Policy Search algorithms,
that compute an approximately optimal policy by following the standard Policy
Iteration (PI) scheme via an -approximate greedy operator (Kakade and Langford,
2002; Lazaric et al., 2010). We describe existing and a few new performance
bounds for Direct Policy Iteration (DPI) (Lagoudakis and Parr, 2003; Fern et
al., 2006; Lazaric et al., 2010) and Conservative Policy Iteration (CPI)
(Kakade and Langford, 2002). By paying a particular attention to the
concentrability constants involved in such guarantees, we notably argue that
the guarantee of CPI is much better than that of DPI, but this comes at the
cost of a relative--exponential in $\frac{1}{\epsilon}$-- increase of time
complexity. We then describe an algorithm, Non-Stationary Direct Policy
Iteration (NSDPI), that can either be seen as 1) a variation of Policy Search
by Dynamic Programming by Bagnell et al. (2003) to the infinite horizon
situation or 2) a simplified version of the Non-Stationary PI with growing
period of Scherrer and Lesner (2012). We provide an analysis of this algorithm,
that shows in particular that it enjoys the best of both worlds: its
performance guarantee is similar to that of CPI, but within a time complexity
similar to that of DPI.
|
Identifying Pairs in Simulated Bio-Medical Time-Series | The paper presents a time-series-based classification approach to identify
similarities in pairs of simulated human-generated patterns. An example for a
pattern is a time-series representing a heart rate during a specific
time-range, wherein the time-series is a sequence of data points that represent
the changes in the heart rate values. A bio-medical simulator system was
developed to acquire a collection of 7,871 price patterns of financial
instruments. The financial instruments traded in real-time on three American
stock exchanges, NASDAQ, NYSE, and AMEX, simulate bio-medical measurements. The
system simulates a human in which each price pattern represents one bio-medical
sensor. Data provided during trading hours from the stock exchanges allowed
real-time classification. Classification is based on new machine learning
techniques: self-labeling, which allows the application of supervised learning
methods on unlabeled time-series and similarity ranking, which applied on a
decision tree learning algorithm to classify time-series regardless of type and
quantity.
|
Predicting Parameters in Deep Learning | We demonstrate that there is significant redundancy in the parameterization
of several deep learning models. Given only a few weight values for each
feature it is possible to accurately predict the remaining values. Moreover, we
show that not only can the parameter values be predicted, but many of them need
not be learned at all. We train several different architectures by learning
only a small number of weights and predicting the rest. In the best case we are
able to predict more than 95% of the weights of a network without any drop in
accuracy.
|
Distributed k-Means and k-Median Clustering on General Topologies | This paper provides new algorithms for distributed clustering for two popular
center-based objectives, k-median and k-means. These algorithms have provable
guarantees and improve communication complexity over existing approaches.
Following a classic approach in clustering by \cite{har2004coresets}, we reduce
the problem of finding a clustering with low cost to the problem of finding a
coreset of small size. We provide a distributed method for constructing a
global coreset which improves over the previous methods by reducing the
communication complexity, and which works over general communication
topologies. Experimental results on large scale data sets show that this
approach outperforms other coreset-based distributed clustering algorithms.
|
Prediction with Missing Data via Bayesian Additive Regression Trees | We present a method for incorporating missing data in non-parametric
statistical learning without the need for imputation. We focus on a tree-based
method, Bayesian Additive Regression Trees (BART), enhanced with "Missingness
Incorporated in Attributes," an approach recently proposed incorporating
missingness into decision trees (Twala, 2008). This procedure takes advantage
of the partitioning mechanisms found in tree-based models. Simulations on
generated models and real data indicate that our proposed method can forecast
well on complicated missing-at-random and not-missing-at-random models as well
as models where missingness itself influences the response. Our procedure has
higher predictive performance and is more stable than competitors in many
cases. We also illustrate BART's abilities to incorporate missingness into
uncertainty intervals and to detect the influence of missingness on the model
fit.
|
Provable Inductive Matrix Completion | Consider a movie recommendation system where apart from the ratings
information, side information such as user's age or movie's genre is also
available. Unlike standard matrix completion, in this setting one should be
able to predict inductively on new users/movies. In this paper, we study the
problem of inductive matrix completion in the exact recovery setting. That is,
we assume that the ratings matrix is generated by applying feature vectors to a
low-rank matrix and the goal is to recover back the underlying matrix.
Furthermore, we generalize the problem to that of low-rank matrix estimation
using rank-1 measurements. We study this generic problem and provide conditions
that the set of measurements should satisfy so that the alternating
minimization method (which otherwise is a non-convex method with no convergence
guarantees) is able to recover back the {\em exact} underlying low-rank matrix.
In addition to inductive matrix completion, we show that two other low-rank
estimation problems can be studied in our framework: a) general low-rank matrix
sensing using rank-1 measurements, and b) multi-label regression with missing
labels. For both the problems, we provide novel and interesting bounds on the
number of measurements required by alternating minimization to provably
converges to the {\em exact} low-rank matrix. In particular, our analysis for
the general low rank matrix sensing problem significantly improves the required
storage and computational cost than that required by the RIP-based matrix
sensing methods \cite{RechtFP2007}. Finally, we provide empirical validation of
our approach and demonstrate that alternating minimization is able to recover
the true matrix for the above mentioned problems using a small number of
measurements.
|
Online Learning under Delayed Feedback | Online learning with delayed feedback has received increasing attention
recently due to its several applications in distributed, web-based learning
problems. In this paper we provide a systematic study of the topic, and analyze
the effect of delay on the regret of online learning algorithms. Somewhat
surprisingly, it turns out that delay increases the regret in a multiplicative
way in adversarial problems, and in an additive way in stochastic problems. We
give meta-algorithms that transform, in a black-box fashion, algorithms
developed for the non-delayed case into ones that can handle the presence of
delays in the feedback loop. Modifications of the well-known UCB algorithm are
also developed for the bandit problem with delayed feedback, with the advantage
over the meta-algorithms that they can be implemented with lower complexity.
|
Fast Gradient-Based Inference with Continuous Latent Variable Models in
Auxiliary Form | We propose a technique for increasing the efficiency of gradient-based
inference and learning in Bayesian networks with multiple layers of continuous
latent vari- ables. We show that, in many cases, it is possible to express such
models in an auxiliary form, where continuous latent variables are
conditionally deterministic given their parents and a set of independent
auxiliary variables. Variables of mod- els in this auxiliary form have much
larger Markov blankets, leading to significant speedups in gradient-based
inference, e.g. rapid mixing Hybrid Monte Carlo and efficient gradient-based
optimization. The relative efficiency is confirmed in ex- periments.
|
A Gang of Bandits | Multi-armed bandit problems are receiving a great deal of attention because
they adequately formalize the exploration-exploitation trade-offs arising in
several industrially relevant applications, such as online advertisement and,
more generally, recommendation systems. In many cases, however, these
applications have a strong social component, whose integration in the bandit
algorithm could lead to a dramatic performance increase. For instance, we may
want to serve content to a group of users by taking advantage of an underlying
network of social relationships among them. In this paper, we introduce novel
algorithmic approaches to the solution of such networked bandit problems. More
specifically, we design and analyze a global strategy which allocates a bandit
algorithm to each network node (user) and allows it to "share" signals
(contexts and payoffs) with the neghboring nodes. We then derive two more
scalable variants of this strategy based on different ways of clustering the
graph nodes. We experimentally compare the algorithm and its variants to
state-of-the-art methods for contextual bandits that do not use the relational
information. Our experiments, carried out on synthetic and real-world datasets,
show a marked increase in prediction performance obtained by exploiting the
network structure.
|
Kernel Mean Estimation and Stein's Effect | A mean function in reproducing kernel Hilbert space, or a kernel mean, is an
important part of many applications ranging from kernel principal component
analysis to Hilbert-space embedding of distributions. Given finite samples, an
empirical average is the standard estimate for the true kernel mean. We show
that this estimator can be improved via a well-known phenomenon in statistics
called Stein's phenomenon. After consideration, our theoretical analysis
reveals the existence of a wide class of estimators that are better than the
standard. Focusing on a subset of this class, we propose efficient shrinkage
estimators for the kernel mean. Empirical evaluations on several benchmark
applications clearly demonstrate that the proposed estimators outperform the
standard kernel mean estimator.
|
$\propto$SVM for learning with label proportions | We study the problem of learning with label proportions in which the training
data is provided in groups and only the proportion of each class in each group
is known. We propose a new method called proportion-SVM, or $\propto$SVM, which
explicitly models the latent unknown instance labels together with the known
group label proportions in a large-margin framework. Unlike the existing works,
our approach avoids making restrictive assumptions about the data. The
$\propto$SVM model leads to a non-convex integer programming problem. In order
to solve it efficiently, we propose two algorithms: one based on simple
alternating optimization and the other based on a convex relaxation. Extensive
experiments on standard datasets show that $\propto$SVM outperforms the
state-of-the-art, especially for larger group sizes.
|
(More) Efficient Reinforcement Learning via Posterior Sampling | Most provably-efficient learning algorithms introduce optimism about
poorly-understood states and actions to encourage exploration. We study an
alternative approach for efficient exploration, posterior sampling for
reinforcement learning (PSRL). This algorithm proceeds in repeated episodes of
known duration. At the start of each episode, PSRL updates a prior distribution
over Markov decision processes and takes one sample from this posterior. PSRL
then follows the policy that is optimal for this sample during the episode. The
algorithm is conceptually simple, computationally efficient and allows an agent
to encode prior knowledge in a natural way. We establish an $\tilde{O}(\tau S
\sqrt{AT})$ bound on the expected regret, where $T$ is time, $\tau$ is the
episode length and $S$ and $A$ are the cardinalities of the state and action
spaces. This bound is one of the first for an algorithm not based on optimism,
and close to the state of the art for any reinforcement learning algorithm. We
show through simulation that PSRL significantly outperforms existing algorithms
with similar regret bounds.
|
Bayesian Differential Privacy through Posterior Sampling | Differential privacy formalises privacy-preserving mechanisms that provide
access to a database. We pose the question of whether Bayesian inference itself
can be used directly to provide private access to data, with no modification.
The answer is affirmative: under certain conditions on the prior, sampling from
the posterior distribution can be used to achieve a desired level of privacy
and utility. To do so, we generalise differential privacy to arbitrary dataset
metrics, outcome spaces and distribution families. This allows us to also deal
with non-i.i.d or non-tabular datasets. We prove bounds on the sensitivity of
the posterior to the data, which gives a measure of robustness. We also show
how to use posterior sampling to provide differentially private responses to
queries, within a decision-theoretic framework. Finally, we provide bounds on
the utility and on the distinguishability of datasets. The latter are
complemented by a novel use of Le Cam's method to obtain lower bounds. All our
general results hold for arbitrary database metrics, including those for the
common definition of differential privacy. For specific choices of the metric,
we give a number of examples satisfying our assumptions.
|
Discriminative Parameter Estimation for Random Walks Segmentation:
Technical Report | The Random Walks (RW) algorithm is one of the most e - cient and easy-to-use
probabilistic segmentation methods. By combining contrast terms with prior
terms, it provides accurate segmentations of medical images in a fully
automated manner. However, one of the main drawbacks of using the RW algorithm
is that its parameters have to be hand-tuned. we propose a novel discriminative
learning framework that estimates the parameters using a training dataset. The
main challenge we face is that the training samples are not fully supervised.
Speci cally, they provide a hard segmentation of the images, instead of a
proba-bilistic segmentation. We overcome this challenge by treating the optimal
probabilistic segmentation that is compatible with the given hard segmentation
as a latent variable. This allows us to employ the latent support vector
machine formulation for parameter estimation. We show that our approach signi
cantly outperforms the baseline methods on a challenging dataset consisting of
real clinical 3D MRI volumes of skeletal muscles.
|
Deep Generative Stochastic Networks Trainable by Backprop | We introduce a novel training principle for probabilistic models that is an
alternative to maximum likelihood. The proposed Generative Stochastic Networks
(GSN) framework is based on learning the transition operator of a Markov chain
whose stationary distribution estimates the data distribution. The transition
distribution of the Markov chain is conditional on the previous state,
generally involving a small move, so this conditional distribution has fewer
dominant modes, being unimodal in the limit of small moves. Thus, it is easier
to learn because it is easier to approximate its partition function, more like
learning to perform supervised function approximation, with gradients that can
be obtained by backprop. We provide theorems that generalize recent work on the
probabilistic interpretation of denoising autoencoders and obtain along the way
an interesting justification for dependency networks and generalized
pseudolikelihood, along with a definition of an appropriate joint distribution
and sampling mechanism even when the conditionals are not consistent. GSNs can
be used with missing inputs and can be used to sample subsets of variables
given the rest. We validate these theoretical results with experiments on two
image datasets using an architecture that mimics the Deep Boltzmann Machine
Gibbs sampler but allows training to proceed with simple backprop, without the
need for layerwise pretraining.
|
Multiclass Total Variation Clustering | Ideas from the image processing literature have recently motivated a new set
of clustering algorithms that rely on the concept of total variation. While
these algorithms perform well for bi-partitioning tasks, their recursive
extensions yield unimpressive results for multiclass clustering tasks. This
paper presents a general framework for multiclass total variation clustering
that does not rely on recursion. The results greatly outperform previous total
variation algorithms and compare well with state-of-the-art NMF approaches.
|
Multiclass Semi-Supervised Learning on Graphs using Ginzburg-Landau
Functional Minimization | We present a graph-based variational algorithm for classification of
high-dimensional data, generalizing the binary diffuse interface model to the
case of multiple classes. Motivated by total variation techniques, the method
involves minimizing an energy functional made up of three terms. The first two
terms promote a stepwise continuous classification function with sharp
transitions between classes, while preserving symmetry among the class labels.
The third term is a data fidelity term, allowing us to incorporate prior
information into the model in a semi-supervised framework. The performance of
the algorithm on synthetic data, as well as on the COIL and MNIST benchmark
datasets, is competitive with state-of-the-art graph-based multiclass
segmentation methods.
|
Verdict Accuracy of Quick Reduct Algorithm using Clustering and
Classification Techniques for Gene Expression Data | In most gene expression data, the number of training samples is very small
compared to the large number of genes involved in the experiments. However,
among the large amount of genes, only a small fraction is effective for
performing a certain task. Furthermore, a small subset of genes is desirable in
developing gene expression based diagnostic tools for delivering reliable and
understandable results. With the gene selection results, the cost of biological
experiment and decision can be greatly reduced by analyzing only the marker
genes. An important application of gene expression data in functional genomics
is to classify samples according to their gene expression profiles. Feature
selection (FS) is a process which attempts to select more informative features.
It is one of the important steps in knowledge discovery. Conventional
supervised FS methods evaluate various feature subsets using an evaluation
function or metric to select only those features which are related to the
decision classes of the data under consideration. This paper studies a feature
selection method based on rough set theory. Further K-Means, Fuzzy C-Means
(FCM) algorithm have implemented for the reduced feature set without
considering class labels. Then the obtained results are compared with the
original class labels. Back Propagation Network (BPN) has also been used for
classification. Then the performance of K-Means, FCM, and BPN are analyzed
through the confusion matrix. It is found that the BPN is performing well
comparatively.
|
Performance analysis of unsupervised feature selection methods | Feature selection (FS) is a process which attempts to select more informative
features. In some cases, too many redundant or irrelevant features may
overpower main features for classification. Feature selection can remedy this
problem and therefore improve the prediction accuracy and reduce the
computational overhead of classification algorithms. The main aim of feature
selection is to determine a minimal feature subset from a problem domain while
retaining a suitably high accuracy in representing the original features. In
this paper, Principal Component Analysis (PCA), Rough PCA, Unsupervised Quick
Reduct (USQR) algorithm and Empirical Distribution Ranking (EDR) approaches are
applied to discover discriminative features that will be the most adequate ones
for classification. Efficiency of the approaches is evaluated using standard
classification metrics.
|
Diffusion map for clustering fMRI spatial maps extracted by independent
component analysis | Functional magnetic resonance imaging (fMRI) produces data about activity
inside the brain, from which spatial maps can be extracted by independent
component analysis (ICA). In datasets, there are n spatial maps that contain p
voxels. The number of voxels is very high compared to the number of analyzed
spatial maps. Clustering of the spatial maps is usually based on correlation
matrices. This usually works well, although such a similarity matrix inherently
can explain only a certain amount of the total variance contained in the
high-dimensional data where n is relatively small but p is large. For
high-dimensional space, it is reasonable to perform dimensionality reduction
before clustering. In this research, we used the recently developed diffusion
map for dimensionality reduction in conjunction with spectral clustering. This
research revealed that the diffusion map based clustering worked as well as the
more traditional methods, and produced more compact clusters when needed.
|
Tight Lower Bound on the Probability of a Binomial Exceeding its
Expectation | We give the proof of a tight lower bound on the probability that a binomial
random variable exceeds its expected value. The inequality plays an important
role in a variety of contexts, including the analysis of relative deviation
bounds in learning theory and generalization bounds for unbounded loss
functions.
|
Highly Scalable, Parallel and Distributed AdaBoost Algorithm using Light
Weight Threads and Web Services on a Network of Multi-Core Machines | AdaBoost is an important algorithm in machine learning and is being widely
used in object detection. AdaBoost works by iteratively selecting the best
amongst weak classifiers, and then combines several weak classifiers to obtain
a strong classifier. Even though AdaBoost has proven to be very effective, its
learning execution time can be quite large depending upon the application e.g.,
in face detection, the learning time can be several days. Due to its increasing
use in computer vision applications, the learning time needs to be drastically
reduced so that an adaptive near real time object detection system can be
incorporated. In this paper, we develop a hybrid parallel and distributed
AdaBoost algorithm that exploits the multiple cores in a CPU via light weight
threads, and also uses multiple machines via a web service software
architecture to achieve high scalability. We present a novel hierarchical web
services based distributed architecture and achieve nearly linear speedup up to
the number of processors available to us. In comparison with the previously
published work, which used a single level master-slave parallel and distributed
implementation [1] and only achieved a speedup of 2.66 on four nodes, we
achieve a speedup of 95.1 on 31 workstations each having a quad-core processor,
resulting in a learning time of only 4.8 seconds per feature.
|
Gaussian Process-Based Decentralized Data Fusion and Active Sensing for
Mobility-on-Demand System | Mobility-on-demand (MoD) systems have recently emerged as a promising
paradigm of one-way vehicle sharing for sustainable personal urban mobility in
densely populated cities. In this paper, we enhance the capability of a MoD
system by deploying robotic shared vehicles that can autonomously cruise the
streets to be hailed by users. A key challenge to managing the MoD system
effectively is that of real-time, fine-grained mobility demand sensing and
prediction. This paper presents a novel decentralized data fusion and active
sensing algorithm for real-time, fine-grained mobility demand sensing and
prediction with a fleet of autonomous robotic vehicles in a MoD system. Our
Gaussian process (GP)-based decentralized data fusion algorithm can achieve a
fine balance between predictive power and time efficiency. We theoretically
guarantee its predictive performance to be equivalent to that of a
sophisticated centralized sparse approximation for the GP model: The
computation of such a sparse approximate GP model can thus be distributed among
the MoD vehicles, hence achieving efficient and scalable demand prediction.
Though our decentralized active sensing strategy is devised to gather the most
informative demand data for demand prediction, it can achieve a dual effect of
fleet rebalancing to service the mobility demands. Empirical evaluation on
real-world mobility demand data shows that our proposed algorithm can achieve a
better balance between predictive accuracy and time efficiency than
state-of-the-art algorithms.
|
Policy Search: Any Local Optimum Enjoys a Global Performance Guarantee | Local Policy Search is a popular reinforcement learning approach for handling
large state spaces. Formally, it searches locally in a paramet erized policy
space in order to maximize the associated value function averaged over some
predefined distribution. It is probably commonly b elieved that the best one
can hope in general from such an approach is to get a local optimum of this
criterion. In this article, we show th e following surprising result:
\emph{any} (approximate) \emph{local optimum} enjoys a \emph{global performance
guarantee}. We compare this g uarantee with the one that is satisfied by Direct
Policy Iteration, an approximate dynamic programming algorithm that does some
form of Poli cy Search: if the approximation error of Local Policy Search may
generally be bigger (because local search requires to consider a space of s
tochastic policies), we argue that the concentrability coefficient that appears
in the performance bound is much nicer. Finally, we discuss several practical
and theoretical consequences of our analysis.
|
Fast greedy algorithm for subspace clustering from corrupted and
incomplete data | We describe the Fast Greedy Sparse Subspace Clustering (FGSSC) algorithm
providing an efficient method for clustering data belonging to a few
low-dimensional linear or affine subspaces. The main difference of our
algorithm from predecessors is its ability to work with noisy data having a
high rate of erasures (missed entries with the known coordinates) and errors
(corrupted entries with unknown coordinates). We discuss here how to implement
the fast version of the greedy algorithm with the maximum efficiency whose
greedy strategy is incorporated into iterations of the basic algorithm.
We provide numerical evidences that, in the subspace clustering capability,
the fast greedy algorithm outperforms not only the existing state-of-the art
SSC algorithm taken by the authors as a basic algorithm but also the recent
GSSC algorithm. At the same time, its computational cost is only slightly
higher than the cost of SSC.
The numerical evidence of the algorithm significant advantage is presented
for a few synthetic models as well as for the Extended Yale B dataset of facial
images. In particular, the face recognition misclassification rate turned out
to be 6-20 times lower than for the SSC algorithm. We provide also the
numerical evidence that the FGSSC algorithm is able to perform clustering of
corrupted data efficiently even when the sum of subspace dimensions
significantly exceeds the dimension of the ambient space.
|
Loss-Proportional Subsampling for Subsequent ERM | We propose a sampling scheme suitable for reducing a data set prior to
selecting a hypothesis with minimum empirical risk. The sampling only considers
a subset of the ultimate (unknown) hypothesis set, but can nonetheless
guarantee that the final excess risk will compare favorably with utilizing the
entire original data set. We demonstrate the practical benefits of our approach
on a large dataset which we subsample and subsequently fit with boosted trees.
|
Emotional Expression Classification using Time-Series Kernels | Estimation of facial expressions, as spatio-temporal processes, can take
advantage of kernel methods if one considers facial landmark positions and
their motion in 3D space. We applied support vector classification with kernels
derived from dynamic time-warping similarity measures. We achieved over 99%
accuracy - measured by area under ROC curve - using only the 'motion pattern'
of the PCA compressed representation of the marker point vector, the so-called
shape parameters. Beyond the classification of full motion patterns, several
expressions were recognized with over 90% accuracy in as few as 5-6 frames from
their onset, about 200 milliseconds.
|
Minimax Theory for High-dimensional Gaussian Mixtures with Sparse Mean
Separation | While several papers have investigated computationally and statistically
efficient methods for learning Gaussian mixtures, precise minimax bounds for
their statistical performance as well as fundamental limits in high-dimensional
settings are not well-understood. In this paper, we provide precise information
theoretic bounds on the clustering accuracy and sample complexity of learning a
mixture of two isotropic Gaussians in high dimensions under small mean
separation. If there is a sparse subset of relevant dimensions that determine
the mean separation, then the sample complexity only depends on the number of
relevant dimensions and mean separation, and can be achieved by a simple
computationally efficient procedure. Our results provide the first step of a
theoretical basis for recent methods that combine feature selection and
clustering.
|
Logistic Tensor Factorization for Multi-Relational Data | Tensor factorizations have become increasingly popular approaches for various
learning tasks on structured data. In this work, we extend the RESCAL tensor
factorization, which has shown state-of-the-art results for multi-relational
learning, to account for the binary nature of adjacency tensors. We study the
improvements that can be gained via this approach on various benchmark datasets
and show that the logistic extension can improve the prediction results
significantly.
|
Predicting Risk-of-Readmission for Congestive Heart Failure Patients: A
Multi-Layer Approach | Mitigating risk-of-readmission of Congestive Heart Failure (CHF) patients
within 30 days of discharge is important because such readmissions are not only
expensive but also critical indicator of provider care and quality of
treatment. Accurately predicting the risk-of-readmission may allow hospitals to
identify high-risk patients and eventually improve quality of care by
identifying factors that contribute to such readmissions in many scenarios. In
this paper, we investigate the problem of predicting risk-of-readmission as a
supervised learning problem, using a multi-layer classification approach.
Earlier contributions inadequately attempted to assess a risk value for 30 day
readmission by building a direct predictive model as opposed to our approach.
We first split the problem into various stages, (a) at risk in general (b) risk
within 60 days (c) risk within 30 days, and then build suitable classifiers for
each stage, thereby increasing the ability to accurately predict the risk using
multiple layers of decision. The advantage of our approach is that we can use
different classification models for the subtasks that are more suited for the
respective problems. Moreover, each of the subtasks can be solved using
different features and training data leading to a highly confident diagnosis or
risk compared to a one-shot single layer approach. An experimental evaluation
on actual hospital patient record data from Multicare Health Systems shows that
our model is significantly better at predicting risk-of-readmission of CHF
patients within 30 days after discharge compared to prior attempts.
|
A Novel Approach for Single Gene Selection Using Clustering and
Dimensionality Reduction | We extend the standard rough set-based approach to deal with huge amounts of
numeric attributes versus small amount of available objects. Here, a novel
approach of clustering along with dimensionality reduction; Hybrid Fuzzy C
Means-Quick Reduct (FCMQR) algorithm is proposed for single gene selection.
Gene selection is a process to select genes which are more informative. It is
one of the important steps in knowledge discovery. The problem is that all
genes are not important in gene expression data. Some of the genes may be
redundant, and others may be irrelevant and noisy. In this study, the entire
dataset is divided in proper grouping of similar genes by applying Fuzzy C
Means (FCM) algorithm. A high class discriminated genes has been selected based
on their degree of dependence by applying Quick Reduct algorithm based on Rough
Set Theory to all the resultant clusters. Average Correlation Value (ACV) is
calculated for the high class discriminated genes. The clusters which have the
ACV value a s 1 is determined as significant clusters, whose classification
accuracy will be equal or high when comparing to the accuracy of the entire
dataset. The proposed algorithm is evaluated using WEKA classifiers and
compared. Finally, experimental results related to the leukemia cancer data
confirm that our approach is quite promising, though it surely requires further
research.
|
Non-strongly-convex smooth stochastic approximation with convergence
rate O(1/n) | We consider the stochastic approximation problem where a convex function has
to be minimized, given only the knowledge of unbiased estimates of its
gradients at certain points, a framework which includes machine learning
methods based on the minimization of the empirical risk. We focus on problems
without strong convexity, for which all previously known algorithms achieve a
convergence rate for function values of O(1/n^{1/2}). We consider and analyze
two algorithms that achieve a rate of O(1/n) for classical supervised learning
problems. For least-squares regression, we show that averaged stochastic
gradient descent with constant step-size achieves the desired rate. For
logistic regression, this is achieved by a simple novel stochastic gradient
algorithm that (a) constructs successive local quadratic approximations of the
loss functions, while (b) preserving the same running time complexity as
stochastic gradient descent. For these algorithms, we provide a non-asymptotic
analysis of the generalization error (in expectation, and also in high
probability for least-squares), and run extensive experiments on standard
machine learning benchmarks showing that they often outperform existing
approaches.
|
Asymptotically Optimal Sequential Estimation of the Mean Based on
Inclusion Principle | A large class of problems in sciences and engineering can be formulated as
the general problem of constructing random intervals with pre-specified
coverage probabilities for the mean. Wee propose a general approach for
statistical inference of mean values based on accumulated observational data.
We show that the construction of such random intervals can be accomplished by
comparing the endpoints of random intervals with confidence sequences for the
mean. Asymptotic results are obtained for such sequential methods.
|
Markov random fields factorization with context-specific independences | Markov random fields provide a compact representation of joint probability
distributions by representing its independence properties in an undirected
graph. The well-known Hammersley-Clifford theorem uses these conditional
independences to factorize a Gibbs distribution into a set of factors. However,
an important issue of using a graph to represent independences is that it
cannot encode some types of independence relations, such as the
context-specific independences (CSIs). They are a particular case of
conditional independences that is true only for a certain assignment of its
conditioning set; in contrast to conditional independences that must hold for
all its assignments. This work presents a method for factorizing a Markov
random field according to CSIs present in a distribution, and formally
guarantees that this factorization is correct. This is presented in our main
contribution, the context-specific Hammersley-Clifford theorem, a
generalization to CSIs of the Hammersley-Clifford theorem that applies for
conditional independences.
|
Generative Model Selection Using a Scalable and Size-Independent Complex
Network Classifier | Real networks exhibit nontrivial topological features such as heavy-tailed
degree distribution, high clustering, and small-worldness. Researchers have
developed several generative models for synthesizing artificial networks that
are structurally similar to real networks. An important research problem is to
identify the generative model that best fits to a target network. In this
paper, we investigate this problem and our goal is to select the model that is
able to generate graphs similar to a given network instance. By the means of
generating synthetic networks with seven outstanding generative models, we have
utilized machine learning methods to develop a decision tree for model
selection. Our proposed method, which is named "Generative Model Selection for
Complex Networks" (GMSCN), outperforms existing methods with respect to
accuracy, scalability and size-independence.
|
Auditing: Active Learning with Outcome-Dependent Query Costs | We propose a learning setting in which unlabeled data is free, and the cost
of a label depends on its value, which is not known in advance. We study binary
classification in an extreme case, where the algorithm only pays for negative
labels. Our motivation are applications such as fraud detection, in which
investigating an honest transaction should be avoided if possible. We term the
setting auditing, and consider the auditing complexity of an algorithm: the
number of negative labels the algorithm requires in order to learn a hypothesis
with low relative error. We design auditing algorithms for simple hypothesis
classes (thresholds and rectangles), and show that with these algorithms, the
auditing complexity can be significantly lower than the active label
complexity. We also discuss a general competitive approach for auditing and
possible modifications to the framework.
|
DISCOMAX: A Proximity-Preserving Distance Correlation Maximization
Algorithm | In a regression setting we propose algorithms that reduce the dimensionality
of the features while simultaneously maximizing a statistical measure of
dependence known as distance correlation between the low-dimensional features
and a response variable. This helps in solving the prediction problem with a
low-dimensional set of features. Our setting is different from subset-selection
algorithms where the problem is to choose the best subset of features for
regression. Instead, we attempt to generate a new set of low-dimensional
features as in a feature-learning setting. We attempt to keep our proposed
approach as model-free and our algorithm does not assume the application of any
specific regression model in conjunction with the low-dimensional features that
it learns. The algorithm is iterative and is fomulated as a combination of the
majorization-minimization and concave-convex optimization procedures. We also
present spectral radius based convergence results for the proposed iterations.
|
Efficient Classification for Metric Data | Recent advances in large-margin classification of data residing in general
metric spaces (rather than Hilbert spaces) enable classification under various
natural metrics, such as string edit and earthmover distance. A general
framework developed for this purpose by von Luxburg and Bousquet [JMLR, 2004]
left open the questions of computational efficiency and of providing direct
bounds on generalization error.
We design a new algorithm for classification in general metric spaces, whose
runtime and accuracy depend on the doubling dimension of the data points, and
can thus achieve superior classification performance in many common scenarios.
The algorithmic core of our approach is an approximate (rather than exact)
solution to the classical problems of Lipschitz extension and of Nearest
Neighbor Search. The algorithm's generalization performance is guaranteed via
the fat-shattering dimension of Lipschitz classifiers, and we present
experimental evidence of its superiority to some common kernel methods. As a
by-product, we offer a new perspective on the nearest neighbor classifier,
which yields significantly sharper risk asymptotics than the classic analysis
of Cover and Hart [IEEE Trans. Info. Theory, 1967].
|
The association problem in wireless networks: a Policy Gradient
Reinforcement Learning approach | The purpose of this paper is to develop a self-optimized association
algorithm based on PGRL (Policy Gradient Reinforcement Learning), which is both
scalable, stable and robust. The term robust means that performance degradation
in the learning phase should be forbidden or limited to predefined thresholds.
The algorithm is model-free (as opposed to Value Iteration) and robust (as
opposed to Q-Learning). The association problem is modeled as a Markov Decision
Process (MDP). The policy space is parameterized. The parameterized family of
policies is then used as expert knowledge for the PGRL. The PGRL converges
towards a local optimum and the average cost decreases monotonically during the
learning process. The properties of the solution make it a good candidate for
practical implementation. Furthermore, the robustness property allows to use
the PGRL algorithm in an "always-on" learning mode.
|
Concentration bounds for temporal difference learning with linear
function approximation: The case of batch data and uniform sampling | We propose a stochastic approximation (SA) based method with randomization of
samples for policy evaluation using the least squares temporal difference
(LSTD) algorithm. Our proposed scheme is equivalent to running regular temporal
difference learning with linear function approximation, albeit with samples
picked uniformly from a given dataset. Our method results in an $O(d)$
improvement in complexity in comparison to LSTD, where $d$ is the dimension of
the data. We provide non-asymptotic bounds for our proposed method, both in
high probability and in expectation, under the assumption that the matrix
underlying the LSTD solution is positive definite. The latter assumption can be
easily satisfied for the pathwise LSTD variant proposed in [23]. Moreover, we
also establish that using our method in place of LSTD does not impact the rate
of convergence of the approximate value function to the true value function.
These rate results coupled with the low computational complexity of our method
make it attractive for implementation in big data settings, where $d$ is large.
A similar low-complexity alternative for least squares regression is well-known
as the stochastic gradient descent (SGD) algorithm. We provide finite-time
bounds for SGD. We demonstrate the practicality of our method as an efficient
alternative for pathwise LSTD empirically by combining it with the least
squares policy iteration (LSPI) algorithm in a traffic signal control
application. We also conduct another set of experiments that combines the SA
based low-complexity variant for least squares regression with the LinUCB
algorithm for contextual bandits, using the large scale news recommendation
dataset from Yahoo.
|
Large Margin Low Rank Tensor Analysis | Other than vector representations, the direct objects of human cognition are
generally high-order tensors, such as 2D images and 3D textures. From this
fact, two interesting questions naturally arise: How does the human brain
represent these tensor perceptions in a "manifold" way, and how can they be
recognized on the "manifold"? In this paper, we present a supervised model to
learn the intrinsic structure of the tensors embedded in a high dimensional
Euclidean space. With the fixed point continuation procedures, our model
automatically and jointly discovers the optimal dimensionality and the
representations of the low dimensional embeddings. This makes it an effective
simulation of the cognitive process of human brain. Furthermore, the
generalization of our model based on similarity between the learned low
dimensional embeddings can be viewed as counterpart of recognition of human
brain. Experiments on applications for object recognition and face recognition
demonstrate the superiority of our proposed model over state-of-the-art
approaches.
|
Precisely Verifying the Null Space Conditions in Compressed Sensing: A
Sandwiching Algorithm | In this paper, we propose new efficient algorithms to verify the null space
condition in compressed sensing (CS). Given an $(n-m) \times n$ ($m>0$) CS
matrix $A$ and a positive $k$, we are interested in computing $\displaystyle
\alpha_k = \max_{\{z: Az=0,z\neq 0\}}\max_{\{K: |K|\leq k\}}$ ${\|z_K
\|_{1}}{\|z\|_{1}}$, where $K$ represents subsets of $\{1,2,...,n\}$, and $|K|$
is the cardinality of $K$. In particular, we are interested in finding the
maximum $k$ such that $\alpha_k < {1}{2}$. However, computing $\alpha_k$ is
known to be extremely challenging. In this paper, we first propose a series of
new polynomial-time algorithms to compute upper bounds on $\alpha_k$. Based on
these new polynomial-time algorithms, we further design a new sandwiching
algorithm, to compute the \emph{exact} $\alpha_k$ with greatly reduced
complexity. When needed, this new sandwiching algorithm also achieves a smooth
tradeoff between computational complexity and result accuracy. Empirical
results show the performance improvements of our algorithm over existing known
methods; and our algorithm outputs precise values of $\alpha_k$, with much
lower complexity than exhaustive search.
|
R3MC: A Riemannian three-factor algorithm for low-rank matrix completion | We exploit the versatile framework of Riemannian optimization on quotient
manifolds to develop R3MC, a nonlinear conjugate-gradient method for low-rank
matrix completion. The underlying search space of fixed-rank matrices is
endowed with a novel Riemannian metric that is tailored to the least-squares
cost. Numerical comparisons suggest that R3MC robustly outperforms
state-of-the-art algorithms across different problem instances, especially
those that combine scarcely sampled and ill-conditioned data.
|
Flexible sampling of discrete data correlations without the marginal
distributions | Learning the joint dependence of discrete variables is a fundamental problem
in machine learning, with many applications including prediction, clustering
and dimensionality reduction. More recently, the framework of copula modeling
has gained popularity due to its modular parametrization of joint
distributions. Among other properties, copulas provide a recipe for combining
flexible models for univariate marginal distributions with parametric families
suitable for potentially high dimensional dependence structures. More
radically, the extended rank likelihood approach of Hoff (2007) bypasses
learning marginal models completely when such information is ancillary to the
learning task at hand as in, e.g., standard dimensionality reduction problems
or copula parameter estimation. The main idea is to represent data by their
observable rank statistics, ignoring any other information from the marginals.
Inference is typically done in a Bayesian framework with Gaussian copulas, and
it is complicated by the fact this implies sampling within a space where the
number of constraints increases quadratically with the number of data points.
The result is slow mixing when using off-the-shelf Gibbs sampling. We present
an efficient algorithm based on recent advances on constrained Hamiltonian
Markov chain Monte Carlo that is simple to implement and does not require
paying for a quadratic cost in sample size.
|
Copula Mixed-Membership Stochastic Blockmodel for Intra-Subgroup
Correlations | The \emph{Mixed-Membership Stochastic Blockmodel (MMSB)} is a popular
framework for modeling social network relationships. It can fully exploit each
individual node's participation (or membership) in a social structure. Despite
its powerful representations, this model makes an assumption that the
distributions of relational membership indicators between two nodes are
independent. Under many social network settings, however, it is possible that
certain known subgroups of people may have high or low correlations in terms of
their membership categories towards each other, and such prior information
should be incorporated into the model. To this end, we introduce a \emph{Copula
Mixed-Membership Stochastic Blockmodel (cMMSB)} where an individual Copula
function is employed to jointly model the membership pairs of those nodes
within the subgroup of interest. The model enables the use of various Copula
functions to suit the scenario, while maintaining the membership's marginal
distribution, as needed, for modeling membership indicators with other nodes
outside of the subgroup of interest. We describe the proposed model and its
inference algorithm in detail for both the finite and infinite cases. In the
experiment section, we compare our algorithms with other popular models in
terms of link prediction, using both synthetic and real world data.
|
Horizontal and Vertical Ensemble with Deep Representation for
Classification | Representation learning, especially which by using deep learning, has been
widely applied in classification. However, how to use limited size of labeled
data to achieve good classification performance with deep neural network, and
how can the learned features further improve classification remain indefinite.
In this paper, we propose Horizontal Voting Vertical Voting and Horizontal
Stacked Ensemble methods to improve the classification performance of deep
neural networks. In the ICML 2013 Black Box Challenge, via using these methods
independently, Bing Xu achieved 3rd in public leaderboard, and 7th in private
leaderboard; Jingjing Xie achieved 4th in public leaderboard, and 5th in
private leaderboard.
|
Understanding Dropout: Training Multi-Layer Perceptrons with Auxiliary
Independent Stochastic Neurons | In this paper, a simple, general method of adding auxiliary stochastic
neurons to a multi-layer perceptron is proposed. It is shown that the proposed
method is a generalization of recently successful methods of dropout (Hinton et
al., 2012), explicit noise injection (Vincent et al., 2010; Bishop, 1995) and
semantic hashing (Salakhutdinov & Hinton, 2009). Under the proposed framework,
an extension of dropout which allows using separate dropping probabilities for
different hidden neurons, or layers, is found to be available. The use of
different dropping probabilities for hidden layers separately is empirically
investigated.
|
Bayesian Inference and Learning in Gaussian Process State-Space Models
with Particle MCMC | State-space models are successfully used in many areas of science,
engineering and economics to model time series and dynamical systems. We
present a fully Bayesian approach to inference \emph{and learning} (i.e. state
estimation and system identification) in nonlinear nonparametric state-space
models. We place a Gaussian process prior over the state transition dynamics,
resulting in a flexible model able to capture complex dynamical phenomena. To
enable efficient inference, we marginalize over the transition dynamics
function and infer directly the joint smoothing distribution using specially
tailored Particle Markov Chain Monte Carlo samplers. Once a sample from the
smoothing distribution is computed, the state transition predictive
distribution can be formulated analytically. Our approach preserves the full
nonparametric expressivity of the model and can make use of sparse Gaussian
processes to greatly reduce computational complexity.
|
Robust Support Vector Machines for Speaker Verification Task | An important step in speaker verification is extracting features that best
characterize the speaker voice. This paper investigates a front-end processing
that aims at improving the performance of speaker verification based on the
SVMs classifier, in text independent mode. This approach combines features
based on conventional Mel-cepstral Coefficients (MFCCs) and Line Spectral
Frequencies (LSFs) to constitute robust multivariate feature vectors. To reduce
the high dimensionality required for training these feature vectors, we use a
dimension reduction method called principal component analysis (PCA). In order
to evaluate the robustness of these systems, different noisy environments have
been used. The obtained results using TIMIT database showed that, using the
paradigm that combines these spectral cues leads to a significant improvement
in verification accuracy, especially with PCA reduction for low signal-to-noise
ratio noisy environment.
|
Reinforcement learning with restrictions on the action set | Consider a 2-player normal-form game repeated over time. We introduce an
adaptive learning procedure, where the players only observe their own realized
payoff at each stage. We assume that agents do not know their own payoff
function, and have no information on the other player. Furthermore, we assume
that they have restrictions on their own action set such that, at each stage,
their choice is limited to a subset of their action set. We prove that the
empirical distributions of play converge to the set of Nash equilibria for
zero-sum and potential games, and games where one player has two actions.
|
Completing Any Low-rank Matrix, Provably | Matrix completion, i.e., the exact and provable recovery of a low-rank matrix
from a small subset of its elements, is currently only known to be possible if
the matrix satisfies a restrictive structural constraint---known as {\em
incoherence}---on its row and column spaces. In these cases, the subset of
elements is sampled uniformly at random.
In this paper, we show that {\em any} rank-$ r $ $ n$-by-$ n $ matrix can be
exactly recovered from as few as $O(nr \log^2 n)$ randomly chosen elements,
provided this random choice is made according to a {\em specific biased
distribution}: the probability of any element being sampled should be
proportional to the sum of the leverage scores of the corresponding row, and
column. Perhaps equally important, we show that this specific form of sampling
is nearly necessary, in a natural precise sense; this implies that other
perhaps more intuitive sampling schemes fail.
We further establish three ways to use the above result for the setting when
leverage scores are not known \textit{a priori}: (a) a sampling strategy for
the case when only one of the row or column spaces are incoherent, (b) a
two-phase sampling procedure for general matrices that first samples to
estimate leverage scores followed by sampling for exact recovery, and (c) an
analysis showing the advantages of weighted nuclear/trace-norm minimization
over the vanilla un-weighted formulation for the case of non-uniform sampling.
|
Dynamic Infinite Mixed-Membership Stochastic Blockmodel | Directional and pairwise measurements are often used to model
inter-relationships in a social network setting. The Mixed-Membership
Stochastic Blockmodel (MMSB) was a seminal work in this area, and many of its
capabilities were extended since then. In this paper, we propose the
\emph{Dynamic Infinite Mixed-Membership stochastic blockModel (DIM3)}, a
generalised framework that extends the existing work to a potentially infinite
number of communities and mixture memberships for each of the network's nodes.
This model is in a dynamic setting, where additional model parameters are
introduced to reflect the degree of persistence between one's memberships at
consecutive times. Accordingly, two effective posterior sampling strategies and
their results are presented using both synthetic and real data.
|
A Convergence Theorem for the Graph Shift-type Algorithms | Graph Shift (GS) algorithms are recently focused as a promising approach for
discovering dense subgraphs in noisy data. However, there are no theoretical
foundations for proving the convergence of the GS Algorithm. In this paper, we
propose a generic theoretical framework consisting of three key GS components:
simplex of generated sequence set, monotonic and continuous objective function
and closed mapping. We prove that GS algorithms with such components can be
transformed to fit the Zangwill's convergence theorem, and the sequence set
generated by the GS procedures always terminates at a local maximum, or at
worst, contains a subsequence which converges to a local maximum of the
similarity measure function. The framework is verified by expanding it to other
GS-type algorithms and experimental results.
|
Non-parametric Power-law Data Clustering | It has always been a great challenge for clustering algorithms to
automatically determine the cluster numbers according to the distribution of
datasets. Several approaches have been proposed to address this issue,
including the recent promising work which incorporate Bayesian Nonparametrics
into the $k$-means clustering procedure. This approach shows simplicity in
implementation and solidity in theory, while it also provides a feasible way to
inference in large scale datasets. However, several problems remains unsolved
in this pioneering work, including the power-law data applicability, mechanism
to merge centers to avoid the over-fitting problem, clustering order problem,
e.t.c.. To address these issues, the Pitman-Yor Process based k-means (namely
\emph{pyp-means}) is proposed in this paper. Taking advantage of the Pitman-Yor
Process, \emph{pyp-means} treats clusters differently by dynamically and
adaptively changing the threshold to guarantee the generation of power-law
clustering results. Also, one center agglomeration procedure is integrated into
the implementation to be able to merge small but close clusters and then
adaptively determine the cluster number. With more discussion on the clustering
order, the convergence proof, complexity analysis and extension to spectral
clustering, our approach is compared with traditional clustering algorithm and
variational inference methods. The advantages and properties of pyp-means are
validated by experiments on both synthetic datasets and real world datasets.
|
Physeter catodon localization by sparse coding | This paper presents a spermwhale' localization architecture using jointly a
bag-of-features (BoF) approach and machine learning framework. BoF methods are
known, especially in computer vision, to produce from a collection of local
features a global representation invariant to principal signal transformations.
Our idea is to regress supervisely from these local features two rough
estimates of the distance and azimuth thanks to some datasets where both
acoustic events and ground-truth position are now available. Furthermore, these
estimates can feed a particle filter system in order to obtain a precise
spermwhale' position even in mono-hydrophone configuration. Anti-collision
system and whale watching are considered applications of this work.
|
Guaranteed Classification via Regularized Similarity Learning | Learning an appropriate (dis)similarity function from the available data is a
central problem in machine learning, since the success of many machine learning
algorithms critically depends on the choice of a similarity function to compare
examples. Despite many approaches for similarity metric learning have been
proposed, there is little theoretical study on the links between similarity
met- ric learning and the classification performance of the result classifier.
In this paper, we propose a regularized similarity learning formulation
associated with general matrix-norms, and establish their generalization
bounds. We show that the generalization error of the resulting linear separator
can be bounded by the derived generalization bound of similarity learning. This
shows that a good gen- eralization of the learnt similarity function guarantees
a good classification of the resulting linear classifier. Our results extend
and improve those obtained by Bellet at al. [3]. Due to the techniques
dependent on the notion of uniform stability [6], the bound obtained there
holds true only for the Frobenius matrix- norm regularization. Our techniques
using the Rademacher complexity [5] and its related Khinchin-type inequality
enable us to establish bounds for regularized similarity learning formulations
associated with general matrix-norms including sparse L 1 -norm and mixed
(2,1)-norm.
|
Learning Using Privileged Information: SVM+ and Weighted SVM | Prior knowledge can be used to improve predictive performance of learning
algorithms or reduce the amount of data required for training. The same goal is
pursued within the learning using privileged information paradigm which was
recently introduced by Vapnik et al. and is aimed at utilizing additional
information available only at training time -- a framework implemented by SVM+.
We relate the privileged information to importance weighting and show that the
prior knowledge expressible with privileged features can also be encoded by
weights associated with every training example. We show that a weighted SVM can
always replicate an SVM+ solution, while the converse is not true and we
construct a counterexample highlighting the limitations of SVM+. Finally, we
touch on the problem of choosing weights for weighted SVMs when privileged
features are not available.
|
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.