DOCSLIB.ORG

A Kernel Path Algorithm for Support Vector Machines

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
A Kernel Path Algorithm for Support Vector Machines A Kernel Path Algorithm for Support Vector Machines Gang Wang [email protected] Dit-Yan Yeung [email protected] Frederick H. Lochovsky [email protected] Department of Computer Science and Engineering, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong, China Abstract allows the inner product between two points in the fea- The choice of the kernel function which deter- ture space to be computed without having to know the mines the mapping between the input space explicit mapping from the input space to the feature and the feature space is of crucial importance space. Rather, it is simply a function of two points in to kernel methods. The past few years have the input space. In general, the input space does not seen many efforts in learning either the ker- have to be a vector space and hence structured, non- nel function or the kernel matrix. In this vectorial data can also be handled in the same way. paper, we address this model selection issue For a kernel method to perform well, the kernel func- by learning the hyperparameter of the ker- tion often plays a very crucial role. Rather than choos- nel function for a support vector machine ing the kernel function and setting its hyperparame- (SVM). We trace the solution path with re- ters manually, many attempts have been made over spect to the kernel hyperparameter without the past few years to automate this process, at least having to train the model multiple times. partially. Ong et al. (2005) shows that the kernel Given a kernel hyperparameter value and the function is a linear combination of a finite number of optimal solution obtained for that value, we prespecified hyperkernel evaluations, and introduces a find that the solutions of the neighborhood method to learn the kernel function directly in an in- hyperparameters can be calculated exactly. ductive setting. Some approaches have been proposed However, the solution path does not exhibit (Cristianini et al., 2002; Bach et al., 2004; Lanckriet piecewise linearity and extends nonlinearly. et al., 2004; Sonnenburg et al., 2006; Zhang et al., As a result, the breakpoints cannot be com- 2006) to seek a kernel matrix directly or learn a conic puted in advance. We propose a method to combination of kernel matrices from data. approximate the breakpoints. Our method is both efficient and general in the sense that Our paper adopts the kernel function learning ap- it can be applied to many kernel functions in proach. However, unlike the method of hyperkernels, common use. we seek to learn the optimal hyperparameter value for a prespecified kernel function. The traditional approach to this model selection problem is to ap- m 1. Introduction ply methods like -fold cross validation to determine the best choice among a number of prespecified can- Kernel methods (M¨uller et al., 2001; Sch¨olkopf & didate hyperparameter values. Extensive exploration Smola, 2002) have demonstrated great success in solv- such as performing line search for one hyperparameter ing many machine learning and pattern recognition or grid search for two hyperparameters is usually ap- problems. These methods implicitly map data points plied. However, this requires training the model mul- from the input space to some feature space where even tiple times with different hyperparameter values and relatively simple algorithms such as linear methods can hence is computationally prohibitive especially when deliver very impressive performance. The implicit fea- the number of candidate values is large. Keerthi et al. ture mapping is determined by a kernel function, which (2006) proposed a hyperparameter tuning approach th based on minimizing a smooth performance validation Appearing in Proceedings of the 24 International Confer- function. The direction to update the hyperparame- ence on Machine Learning, Corvallis, OR, 2007. Copyright 2007 by the author(s)/owner(s). ters is the (negative) gradient of the validation func- A Kernel Path Algorithm for Support Vector Machines tion with respect to the hyperparameters. This ap- over, our algorithm is general in the sense that it can proach also requires training the model and computing be applied to many kernel functions in common use. the gradient multiple times. Some approaches have been proposed to overcome 2. Problem Formulation these problems. A promising recent approach is based In a binary classification problem, we have a set of on solution path algorithms, which can trace the en- n d training pairs {(xi,yi)} where xi ∈X ⊆R is the tire solution path as a function of the hyperparame- i=1 input data, X is the input space, and yi ∈{−1, +1} is ter without having to train the model multiple times. the class label. The goal is to learn a decision function A solution path algorithm is also called a regulariza- f(x) that can classify the unseen data as accurately tion path algorithm if the hyperparameter involved in as possible. Its associated classifier is sign[f(x)]. Us- path tracing is the regularization parameter in the op- ing the hinge loss, the primal optimization problem for timization problem. Efron et al. (2004) proposed an SVC can be formulated based on the standard regu- algorithm called the least angle regression (LARS). It larization framework as follows: can be applied to trace the regularization path for lin- n λ 2 ear least square regression problem regularized with min Rprimal = i=1 ξi + fK (1) f∈H 2 the L1 norm. The path is piecewise linear and hence . .yf ≥ − ξ it is efficient to explore the entire path just by mon- s t i (xi) 1 i (2) itoring the breakpoints between the linear segments ξi ≥ 0. (3) only. Zhu et al. (2003) proposed an algorithm to com- pute the entire regularization path for the L1-norm Here and below, i =1, ..., n. ·K denotes the norm in support vector classification (SVC) and Hastie et al. the reproducing kernel Hilbert space (RKHS) H corre- (2004) proposed one for the standard L2-norm SVC. sponding to a positive definite kernel function K. λ is They are again based on the property that the paths the regularization parameter which gives the balance are piecewise linear. More generally, Rosset and Zhu between the two opposing components of the objec- (2003) showed that any model with an L1 regulariza- tive function. A kernel function, i.e., Kγ (xi, xj), is a tion and a quadratic, piecewise quadratic, piecewise bivariate function with its two independent variables linear, or linear loss function has a piecewise linear defined in the input space. Here we explicitly show regularization path and hence the entire path can be the hyperparameter γ of the kernel function in the computed efficiently. Bach et al. (2005) explored a subscript. Different values of γ embed the data into nonlinear regularization path for multiple kernel learn- different feature spaces. In SVC, both λ and γ are ing regularized with a block L1-norm. Rosset (2004) hyperparameters in the model. proposed a general path following algorithm to approx- Using Wahba’s representer theorem and Wolfe’s dual- imate the nonlinear regularization path. Besides the ity theorem, we can derive the dual form of the opti- regularization parameter, we showed in our previous mization problem as: work (Wang et al., 2006) that this approach can also n be applied to explore the solution paths for some other max Rdual (β,β0)=λ i=1 βi − (4) β,β model hyperparameters. 0 1 n i,j=1 βiβjyiyjKγ (xi, xj) In this paper, we propose a novel method that traces 2 . n y β the solution path with respect to the kernel hyperpa- s t i=1 i i = 0 (5) rameter in SVC. We refer to this solution path as a 0 ≤ βi ≤ 1, (6) kernel path. Given a kernel hyperparameter value and where the dual variables (Lagrange multipliers) β are the optimal solution obtained for that value, we find introduced for constraint (2). The decision function is that the solutions of the neighborhood hyperparame- given by ters can be calculated exactly. Since the kernel hyper- n parameter is embedded into each entry of the kernel f 1 β y K , β . matrix, the path is piecewise smooth but not piecewise (x)=λ i i γ (x xi)+ 0 (7) i=1 linear. The implication is that the next breakpoint cannot be computed in advance before reaching it like f(x) is expressed as an expansion in terms of only a what other solution path algorithms do. We propose subset of data points, called support vectors (SV), for a method to approximate the breakpoints. Unlike the which βi is nonzero. path following algorithm of Rosset (2004) for nonlin- From the KKT optimality conditions, we have three ear regularization paths, our approach is more efficient cases to consider depending on the value of yif(xi), so that the kernel path can be traced efficiently. More- which in turn determines how much the loss ξi is: A Kernel Path Algorithm for Support Vector Machines • If yif(xi) > 1, then ξi =0andβi =0; hold, then the corresponding linear system becomes a L(β (μ + ),μ+ )=0. • y f ξ β ∈ , E If i (xi) = 1, then i =0and i [0 1]; a Expanding L(βE (μ + ),μ+ ) via first-order Taylor a • If yif(xi) < 1, then ξi > 0andβi =1. series approximation around βE (μ), we have βa μ ,μ βa μ ,μ These three cases refer to points lying outside, at and L( E ( + ) + ) L( E ( ) + ) (13) a inside the margins, respectively. We can see that βi ∂L(βE (μ),μ+ ) a a + a [βE (μ + ) − βE (μ)] . can take non-extreme values other than 0 and 1 at ∂βE the optimal solution βˆ only if the value of yif(xi)is a equal to 1.
Load more

Recommended publications
  • Large-Scale Kernel Ranksvm
    Large-scale Kernel RankSVM Tzu-Ming Kuo∗ Ching-Pei Leey Chih-Jen Linz Abstract proposed [11, 23, 5, 1, 14]. However, for some tasks that Learning to rank is an important task for recommendation the feature set is not rich enough, nonlinear methods systems, online advertisement and web search. Among may be needed. Therefore, it is important to develop those learning to rank methods, rankSVM is a widely efficient training methods for large kernel rankSVM. used model. Both linear and nonlinear (kernel) rankSVM Assume we are given a set of training label-query- have been extensively studied, but the lengthy training instance tuples (yi; qi; xi); yi 2 R; qi 2 S ⊂ Z; xi 2 n time of kernel rankSVM remains a challenging issue. In R ; i = 1; : : : ; l, where S is the set of queries. By this paper, after discussing difficulties of training kernel defining the set of preference pairs as rankSVM, we propose an efficient method to handle these (1.1) P ≡ f(i; j) j q = q ; y > y g with p ≡ jP j; problems. The idea is to reduce the number of variables from i j i j quadratic to linear with respect to the number of training rankSVM [10] solves instances, and efficiently evaluate the pairwise losses. Our setting is applicable to a variety of loss functions. Further, 1 T X min w w + C ξi;j general optimization methods can be easily applied to solve w;ξ 2 (i;j)2P the reformulated problem. Implementation issues are also (1.2) subject to wT (φ (x ) − φ (x )) ≥ 1 − ξ ; carefully considered.
    [Show full text]
  • A Kernel Method for Multi-Labelled Classification
    A kernel method for multi-labelled classification Andre´ Elisseeff and Jason Weston BIOwulf Technologies, 305 Broadway, New York, NY 10007 andre,jason ¡ @barhilltechnologies.com Abstract This article presents a Support Vector Machine (SVM) like learning sys- tem to handle multi-label problems. Such problems are usually decom- posed into many two-class problems but the expressive power of such a system can be weak [5, 7]. We explore a new direct approach. It is based on a large margin ranking system that shares a lot of common proper- ties with SVMs. We tested it on a Yeast gene functional classification problem with positive results. 1 Introduction Many problems in Text Mining or Bioinformatics are multi-labelled. That is, each point in a learning set is associated to a set of labels. Consider for instance the classification task of determining the subjects of a document, or of relating one protein to its many effects on a cell. In either case, the learning task would be to output a set of labels whose size is not known in advance: one document can for instance be about food, meat and finance, although another one would concern only food and fat. Two-class and multi-class classification or ordinal regression problems can all be cast into multi-label ones. This makes the latter quite attractive but at the same time it gives a warning: their generality hides their difficulty to solve them. The number of publications is not going to contradict this statement: we are aware of only a few works about the subject [4, 5, 7] and they all concern text mining applications.
    [Show full text]
  • Density Based Data Clustering
    California State University, San Bernardino CSUSB ScholarWorks Electronic Theses, Projects, and Dissertations Office of aduateGr Studies 3-2015 Density Based Data Clustering Rayan Albarakati California State University - San Bernardino Follow this and additional works at: https://scholarworks.lib.csusb.edu/etd Part of the Other Computer Engineering Commons Recommended Citation Albarakati, Rayan, "Density Based Data Clustering" (2015). Electronic Theses, Projects, and Dissertations. 134. https://scholarworks.lib.csusb.edu/etd/134 This Project is brought to you for free and open access by the Office of aduateGr Studies at CSUSB ScholarWorks. It has been accepted for inclusion in Electronic Theses, Projects, and Dissertations by an authorized administrator of CSUSB ScholarWorks. For more information, please contact [email protected]. California State University, San Bernardino CSUSB ScholarWorks Electronic Theses, Projects, and Dissertations Office of Graduate Studies 3-2015 Density Based Data Clustering Rayan Albarakati Follow this and additional works at: http://scholarworks.lib.csusb.edu/etd This Project is brought to you for free and open access by the Office of Graduate Studies at CSUSB ScholarWorks. It has been accepted for inclusion in Electronic Theses, Projects, and Dissertations by an authorized administrator of CSUSB ScholarWorks. For more information, please contact [email protected], [email protected]. DESNITY BASED DATA CLUSTERING A Project Presented to the Faculty of California State University, San Bernardino In Partial Fulfillment of the Requirements for the Degree Master of Science in Computer Science by Rayan Albarakati March 2015 DESNITY BASED DATA CLUSTERING A Project Presented to the Faculty of California State University, San Bernardino by Rayan Albarakati March 2015 Approved by: Haiyan Qiao, Advisor, School of Computer Date Science and Engineering Owen J.Murphy Krestin Voigt © 2015 Rayan Albarakati ABSTRACT Data clustering is a data analysis technique that groups data based on a measure of similarity.
    [Show full text]
  • A User's Guide to Support Vector Machines
    A User's Guide to Support Vector Machines Asa Ben-Hur Jason Weston Department of Computer Science NEC Labs America Colorado State University Princeton, NJ 08540 USA Abstract The Support Vector Machine (SVM) is a widely used classifier. And yet, obtaining the best results with SVMs requires an understanding of their workings and the various ways a user can influence their accuracy. We provide the user with a basic understanding of the theory behind SVMs and focus on their use in practice. We describe the effect of the SVM parameters on the resulting classifier, how to select good values for those parameters, data normalization, factors that affect training time, and software for training SVMs. 1 Introduction The Support Vector Machine (SVM) is a state-of-the-art classification method introduced in 1992 by Boser, Guyon, and Vapnik [1]. The SVM classifier is widely used in bioinformatics (and other disciplines) due to its high accuracy, ability to deal with high-dimensional data such as gene ex- pression, and flexibility in modeling diverse sources of data [2]. SVMs belong to the general category of kernel methods [4, 5]. A kernel method is an algorithm that depends on the data only through dot-products. When this is the case, the dot product can be replaced by a kernel function which computes a dot product in some possibly high dimensional feature space. This has two advantages: First, the ability to generate non-linear decision boundaries using methods designed for linear classifiers. Second, the use of kernel functions allows the user to apply a classifier to data that have no obvious fixed-dimensional vector space representation.
    [Show full text]
  • The Forgetron: a Kernel-Based Perceptron on a Fixed Budget
    The Forgetron: A Kernel-Based Perceptron on a Fixed Budget Ofer Dekel Shai Shalev-Shwartz Yoram Singer School of Computer Science & Engineering The Hebrew University, Jerusalem 91904, Israel oferd,shais,singer @cs.huji.ac.il { } Abstract The Perceptron algorithm, despite its simplicity, often performs well on online classification problems. The Perceptron becomes especially effec- tive when it is used in conjunction with kernels. However, a common dif- ficulty encountered when implementing kernel-based online algorithms is the amount of memory required to store the online hypothesis, which may grow unboundedly. In this paper we describe and analyze a new in- frastructure for kernel-based learning with the Perceptron while adhering to a strict limit on the number of examples that can be stored. We first describe a template algorithm, called the Forgetron, for online learning on a fixed budget. We then provide specific algorithms and derive a uni- fied mistake bound for all of them. To our knowledge, this is the first online learning paradigm which, on one hand, maintains a strict limit on the number of examples it can store and, on the other hand, entertains a relative mistake bound. We also present experiments with real datasets which underscore the merits of our approach. 1 Introduction The introduction of the Support Vector Machine (SVM) [7] sparked a widespread interest in kernel methods as a means of solving (binary) classification problems. Although SVM was initially stated as a batch-learning technique, it significantly influenced the develop- ment of kernel methods in the online-learning setting. Online classification algorithms that can incorporate kernels include the Perceptron [6], ROMMA [5], ALMA [3], NORMA [4] and the Passive-Aggressive family of algorithms [1].
    [Show full text]
  • Kernel Methods Through the Roof: Handling Billions of Points Efficiently
    Kernel methods through the roof: handling billions of points efficiently Giacomo Meanti Luigi Carratino MaLGa, DIBRIS MaLGa, DIBRIS Università degli Studi di Genova Università degli Studi di Genova [email protected] [email protected] Lorenzo Rosasco Alessandro Rudi MaLGa, DIBRIS, IIT & MIT INRIA - École Normale Supérieure Università degli Studi di Genova PSL Research University [email protected] [email protected] Abstract Kernel methods provide an elegant and principled approach to nonparametric learning, but so far could hardly be used in large scale problems, since naïve imple- mentations scale poorly with data size. Recent advances have shown the benefits of a number of algorithmic ideas, for example combining optimization, numerical linear algebra and random projections. Here, we push these efforts further to develop and test a solver that takes full advantage of GPU hardware. Towards this end, we designed a preconditioned gradient solver for kernel methods exploiting both GPU acceleration and parallelization with multiple GPUs, implementing out-of-core variants of common linear algebra operations to guarantee optimal hardware utilization. Further, we optimize the numerical precision of different operations and maximize efficiency of matrix-vector multiplications. As a result we can experimentally show dramatic speedups on datasets with billions of points, while still guaranteeing state of the art performance. Additionally, we make our software available as an easy to use library1. 1 Introduction Kernel methods provide non-linear/non-parametric extensions of many classical linear models in machine learning and statistics [45, 49]. The data are embedded via a non-linear map into a high dimensional feature space, so that linear models in such a space effectively define non-linear models in the original space.
    [Show full text]
  • An Algorithmic Introduction to Clustering
    AN ALGORITHMIC INTRODUCTION TO CLUSTERING APREPRINT Bernardo Gonzalez Department of Computer Science and Engineering UCSC [email protected] June 11, 2020 The purpose of this document is to provide an easy introductory guide to clustering algorithms. Basic knowledge and exposure to probability (random variables, conditional probability, Bayes’ theorem, independence, Gaussian distribution), matrix calculus (matrix and vector derivatives), linear algebra and analysis of algorithm (specifically, time complexity analysis) is assumed. Starting with Gaussian Mixture Models (GMM), this guide will visit different algorithms like the well-known k-means, DBSCAN and Spectral Clustering (SC) algorithms, in a connected and hopefully understandable way for the reader. The first three sections (Introduction, GMM and k-means) are based on [1]. The fourth section (SC) is based on [2] and [5]. Fifth section (DBSCAN) is based on [6]. Sixth section (Mean Shift) is, to the best of author’s knowledge, original work. Seventh section is dedicated to conclusions and future work. Traditionally, these five algorithms are considered completely unrelated and they are considered members of different families of clustering algorithms: • Model-based algorithms: the data are viewed as coming from a mixture of probability distributions, each of which represents a different cluster. A Gaussian Mixture Model is considered a member of this family • Centroid-based algorithms: any data point in a cluster is represented by the central vector of that cluster, which need not be a part of the dataset taken. k-means is considered a member of this family • Graph-based algorithms: the data is represented using a graph, and the clustering procedure leverage Graph theory tools to create a partition of this graph.
    [Show full text]
  • Knowledge-Based Systems Layer-Constrained Variational
    Knowledge-Based Systems 196 (2020) 105753 Contents lists available at ScienceDirect Knowledge-Based Systems journal homepage: www.elsevier.com/locate/knosys Layer-constrained variational autoencoding kernel density estimation model for anomaly detectionI ∗ Peng Lv b, Yanwei Yu a,b, , Yangyang Fan c, Xianfeng Tang d, Xiangrong Tong b a Department of Computer Science and Technology, Ocean University of China, Qingdao, Shandong 266100, China b School of Computer and Control Engineering, Yantai University, Yantai, Shandong 264005, China c Shanghai Key Lab of Advanced High-Temperature Materials and Precision Forming, Shanghai Jiao Tong University, Shanghai 200240, China d College of Information Sciences and Technology, The Pennsylvania State University, University Park, PA 16802, USA article info a b s t r a c t Article history: Unsupervised techniques typically rely on the probability density distribution of the data to detect Received 16 October 2019 anomalies, where objects with low probability density are considered to be abnormal. However, Received in revised form 5 March 2020 modeling the density distribution of high dimensional data is known to be hard, making the problem of Accepted 7 March 2020 detecting anomalies from high-dimensional data challenging. The state-of-the-art methods solve this Available online 10 March 2020 problem by first applying dimension reduction techniques to the data and then detecting anomalies Keywords: in the low dimensional space. Unfortunately, the low dimensional space does not necessarily preserve Anomaly detection the density distribution of the original high dimensional data. This jeopardizes the effectiveness of Variational autoencoder anomaly detection. In this work, we propose a novel high dimensional anomaly detection method Kernel density estimation called LAKE.
    [Show full text]
  • NS-DBSCAN: a Density-Based Clustering Algorithm in Network Space
    International Journal of Geo-Information Article NS-DBSCAN: A Density-Based Clustering Algorithm in Network Space Tianfu Wang 1, Chang Ren 2 , Yun Luo 1 and Jing Tian 1,3,4,* 1 School of Resource and Environmental Science, Wuhan University, 129 Luoyu Road, Wuhan 430079, China; [email protected] (T.W.); [email protected] (Y.L.) 2 State Key Laboratory of Information Engineering in Surveying Mapping and Remote Sensing, Wuhan University, Wuhan 430079, China; [email protected] 3 Key Laboratory of Geographic Information System, Ministry of Education, Wuhan University, 129 Luoyu Road, Wuhan 430079, China 4 Key Laboratory of Digital Mapping and Land Information Application Engineering, National Administration of Surveying, Mapping and Geoinformation, Wuhan 430079, China * Correspondence: [email protected]; Tel.: +86-1362-863-6229 Received: 27 March 2019; Accepted: 5 May 2019; Published: 8 May 2019 Abstract: Spatial clustering analysis is an important spatial data mining technique. It divides objects into clusters according to their similarities in both location and attribute aspects. It plays an essential role in density distribution identification, hot-spot detection, and trend discovery. Spatial clustering algorithms in the Euclidean space are relatively mature, while those in the network space are less well researched. This study aimed to present a well-known clustering algorithm, named density-based spatial clustering of applications with noise (DBSCAN), to network space and proposed a new clustering algorithm named network space DBSCAN (NS-DBSCAN). Basically, the NS-DBSCAN algorithm used a strategy similar to the DBSCAN algorithm. Furthermore, it provided a new technique for visualizing the density distribution and indicating the intrinsic clustering structure.
    [Show full text]
  • Accountable Off-Policy Evaluation with Kernel Bellman Statistics
    Accountable Off-Policy Evaluation With Kernel Bellman Statistics Yihao Feng 1 Tongzheng Ren * 1 Ziyang Tang * 1 Qiang Liu 1 Abstract experimental data or build high-quality simulators, such We consider off-policy evaluation (OPE), which as robotics, advertisement, online education, and medical evaluates the performance of a new policy from treatment (e.g., Murphy et al., 2001; Li et al., 2011; Hirano observed data collected from previous experi- et al., 2003; Liao et al., 2019). ments, without requiring the execution of the new This work concerns the problem of off-policy evaluation policy. This finds important applications in areas (OPE), which estimates the expected reward of a given tar- with high execution cost or safety concerns, such get policy from off-policy data. Because the information as medical diagnosis, recommendation systems in off-policy data is often limited, typical point estimation and robotics. In practice, due to the limited infor- may suffer from large error in difficult cases. Therefore, for mation from off-policy data, it is highly desirable making high-stakes decisions in areas such as medical treat- to construct rigorous confidence intervals, not just ment, it is crucially important to provide reliable confidence point estimation, for the policy performance. In bounds for quantifying the uncertainties in the estimation. this work, we propose a new variational frame- work which reduces the problem of calculating Importance sampling (IS) provides a basic principle for tight confidence bounds in OPE into an optimiza- developing OPE methods (e.g., Dud´ık et al., 2011; Jiang tion problem on a feasible set that catches the true & Li, 2016; Thomas & Brunskill, 2016; Liu et al., 2018; state-action value function with high probability.
    [Show full text]
  • Information Flows of Diverse Autoencoders
    entropy Article Information Flows of Diverse Autoencoders Sungyeop Lee 1,* and Junghyo Jo 2,3,* 1 Department of Physics and Astronomy, Seoul National University, Seoul 08826, Korea 2 Department of Physics Education and Center for Theoretical Physics and Artificial Intelligence Institute, Seoul National University, Seoul 08826, Korea 3 School of Computational Sciences, Korea Institute for Advanced Study, Seoul 02455, Korea * Correspondence: [email protected] (S.L.); [email protected] (J.J.) Abstract: Deep learning methods have had outstanding performances in various fields. A fundamen- tal query is why they are so effective. Information theory provides a potential answer by interpreting the learning process as the information transmission and compression of data. The information flows can be visualized on the information plane of the mutual information among the input, hidden, and output layers. In this study, we examine how the information flows are shaped by the network parameters, such as depth, sparsity, weight constraints, and hidden representations. Here, we adopt autoencoders as models of deep learning, because (i) they have clear guidelines for their information flows, and (ii) they have various species, such as vanilla, sparse, tied, variational, and label autoen- coders. We measured their information flows using Rényi’s matrix-based a-order entropy functional. As learning progresses, they show a typical fitting phase where the amounts of input-to-hidden and hidden-to-output mutual information both increase. In the last stage of learning, however, some autoencoders show a simplifying phase, previously called the “compression phase”, where input-to-hidden mutual information diminishes. In particular, the sparsity regularization of hidden activities amplifies the simplifying phase.
    [Show full text]
  • Learning a Robust Relevance Model for Search Using Kernel Methods
    JournalofMachineLearningResearch12(2011)1429-1458 Submitted 6/10; Revised 12/10; Published 5/11 Learning a Robust Relevance Model for Search Using Kernel Methods Wei Wu∗ [email protected] MOE -Microsoft Key Laboratory of Statistics and Information Technology Department of Probability and Statistics Peking University No.5 Yiheyuan Road, Haidian District, Beijing, 100871, P. R. China Jun Xu [email protected] Hang Li [email protected] Microsoft Research Asia 13F Building 2 No. 5 Danling Street, Haidian District, Beijing, 100080, P.R. China Satoshi Oyama ∗ [email protected] Graduate School of Information Science and Technology Hokkaido University Kita 14, Nishi 9, Kita-ku, 060-0814, Japan Editor: Corinna Cortes Abstract This paper points out that many search relevance models in information retrieval, such as the Vector Space Model, BM25 and Language Models for Information Retrieval, can be viewed as a similarity function between pairs of objects of different types, referred to as an S-function. An S-function is specifically defined as the dot product between the images of two objects in a Hilbert space mapped from two different input spaces. One advantage of taking this view is that one can take a unified and principled approach to address the issues with regard to search relevance. The paper then proposes employing a kernel method to learn a robust relevance model as an S-function, which can effectively deal with the term mismatch problem, one of the biggest challenges in search. The kernel method exploits a positive semi-definite kernel referred to as an S-kernel. The paper shows that when using an S-kernel the model learned by the kernel method is guaranteed to be an S-function.
    [Show full text]