Sparse discriminant analysis based on estimation of posterior probabilities

ABSTRACT

Fisher's linear discriminant analysis (FLDA) is known as a method to find a discriminative feature space for multi-class classification. As a theory of extending FLDA to an ultimate nonlinear form, optimal nonlinear discriminant analysis (ONDA) has been proposed. ONDA indicates that the best theoretical nonlinear map for maximizing the Fisher's discriminant criterion is formulated by using the Bayesian a posterior probabilities. In addition, the theory proves that FLDA is equivalent to ONDA when the Bayesian a posterior probabilities are approximated by linear regression (LR). Due to some limitations of the linear model, there is room to modify FLDA by using stronger approximation/estimation methods. For the purpose of probability estimation, multi-nominal logistic regression (MLR) is more suitable than LR. Along this line, in this paper, we develop a nonlinear discriminant analysis (NDA) in which the posterior probabilities in ONDA are estimated by MLR. In addition, in this paper, we develop a way to introduce sparseness into discriminant analysis. By applying L1 or L2 regularization to LR or MLR, we can incorporate sparseness in FLDA and our NDA to increase generalization performance. The performance of these methods is evaluated by benchmark experiments using last_exam17 standard datasets and a face classification experiment.     

A comparison of the classical and the linear programming approaches to the classification problem in discriminant analysis

Several mathematical programming approaches to the classification problem in discriminant analysis have recently been introduced. This paper empirically compares these newly introduced classification techniques with Fisher's linear discriminant analysis (FLDA), quadratic discriminant analysis (QDA), logit analysis, and several rank-based procedures for a variety of symmetric and skewed distributions. The percent of correctly classified observations by each procedure in a holdout sample indicate that while under some experimental conditions the linear programming approaches compete well with the classical procedures, overall, however, their performance lags behind that of the classical procedures.     

Asymptotic Optimality of Sparse Linear Discriminant Analysis with Arbitrary Number of Classes

Many sparse linear discriminant analysis (LDA) methods have been proposed to overcome the major problems of the classic LDA in high‐dimensional settings. However, the asymptotic optimality results are limited to the case with only two classes. When there are more than two classes, the classification boundary is complicated and no explicit formulas for the classification errors exist. We consider the asymptotic optimality in the high‐dimensional settings for a large family of linear classification rules with arbitrary number of classes. Our main theorem provides easy‐to‐check criteria for the asymptotic optimality of a general classification rule in this family as dimensionality and sample size both go to infinity and the number of classes is arbitrary. We establish the corresponding convergence rates. The general theory is applied to the classic LDA and the extensions of two recently proposed sparse LDA methods to obtain the asymptotic optimality.     

Time Series Classification Based on Spectral Analysis

For time series data with obvious periodicity (e.g., electric motor systems and cardiac monitor) or vague periodicity (e.g., earthquake and explosion, speech, and stock data), frequency-based techniques using the spectral analysis can usually capture the features of the series. By this approach, we are able not only to reduce the data dimensions into frequency domain but also utilize these frequencies by general classification methods such as linear discriminant analysis (LDA) and k-nearest-neighbor (KNN) to classify the time series. This is a combination of two classical approaches. However, there is a difficulty in using LDA and KNN in frequency domain due to excessive dimensions of data. We overcome the obstacle by using Singular Value Decomposition to select essential frequencies. Two data sets are used to illustrate our approach. The classification error rates of our simple approach are comparable to those of several more complicated methods.     

Sequential correction of linear classifiers

In this article, a sequential correction of two linear methods: linear discriminant analysis (LDA) and perceptron is proposed. This correction relies on sequential joining of additional features on which the classifier is trained. These new features are posterior probabilities determined by a basic classification method such as LDA and perceptron. In each step, we add the probabilities obtained on a slightly different data set, because the vector of added probabilities varies at each step. We therefore have many classifiers of the same type trained on slightly different data sets. Four different sequential correction methods are presented based on different combining schemas (e.g. mean rule and product rule). Experimental results on different data sets demonstrate that the improvements are efficient, and that this approach outperforms classical linear methods, providing a significant reduction in the mean classification error rate.     

An Optimal Semiparametric Method for Two‐group Classification

In the classical discriminant analysis, when two multivariate normal distributions with equal variance–covariance matrices are assumed for two groups, the classical linear discriminant function is optimal with respect to maximizing the standardized difference between the means of two groups. However, for a typical case‐control study, the distributional assumption for the case group often needs to be relaxed in practice. Komori et al. (Generalized t ‐statistic for two‐group classification. Biometrics 2015, 71: 404–416) proposed the generalized t ‐statistic to obtain a linear discriminant function, which allows for heterogeneity of case group. Their procedure has an optimality property in the class of consideration. We perform a further study of the problem and show that additional improvement is achievable. The approach we propose does not require a parametric distributional assumption on the case group. We further show that the new estimator is efficient, in that no further improvement is possible to construct the linear discriminant function more efficiently. We conduct simulation studies and real data examples to illustrate the finite sample performance and the gain that it produces in comparison with existing methods.     

Why do we observe misclassification errors smaller than the Bayes error?

In simulation studies for discriminant analysis, misclassification errors are often computed using the Monte Carlo method, by testing a classifier on large samples generated from known populations. Although large samples are expected to behave closely to the underlying distributions, they may not do so in a small interval or region, and thus may lead to unexpected results. We demonstrate with an example that the LDA misclassification error computed via the Monte Carlo method may often be smaller than the Bayes error. We give a rigorous explanation and recommend a method to properly compute misclassification errors.     

Optimal design for classification of functional data

We study the design problem for the optimal classification of functional data. The goal is to select sampling time points so that functional data observed at these time points can be classified accurately. We propose optimal designs that are applicable to either dense or sparse functional data. Using linear discriminant analysis, we formulate our design objectives as explicit functions of the sampling points. We study the theoretical properties of the proposed design objectives and provide a practical implementation. The performance of the proposed design is evaluated through simulations and real data applications. The Canadian Journal of Statistics 48: 285–307; 2020 © 2019 Statistical Society of Canada     

The rank transformation as a method of discrimination with some examples

The procedure of statistical discrimination Is simple in theory but so simple in practice. An observation x₀possibly uiultivariate, is to be classified into one of several populations π₁,…,π_k which have respectively, the density functions f₁(x), ? ? ? , f_k(x). The decision procedure is to evaluate each density function at X₀ to see which function gives the largest value f_i(X₀) , and then to declare that X₀ belongs to the population corresponding to the largest value. If these den-sities can be assumed to be normal with equal covariance matricesthen the decision procedure is known as Fisher’s linear discrimi-nant function (LDF) method. In the case of unequal covariance matrices the procedure is called the quadratic discriminant func-tion (QDF) method. If the densities cannot be assumed to be nor-mal then the LDF and QDF might not perform well. Several different procedures have appeared in the literature which offer discriminant procedures for nonnormal data. However, these pro-cedures are generally difficult to use and are not readily available as canned statistical programs.

Another approach to discriminant analysis is to use some sortof mathematical trans format ion on the samples so that their distribution function is approximately normal, and then use the convenient LDF and QDF methods. One transformation that:applies to all distributions equally well is the rank transformation. The result of this transformation is that a very simple and easy to use procedure is made available. This procedure is quite robust as is evidenced by comparisons of the rank transform results with several published simulation studies.     

Integrating linear discriminant analysis,polynomial basis expansion,and genetic search for two-group classification

We propose a hybrid two-group classification method that integrates linear discriminant analysis, a polynomial expansion of the basis (or variable space), and a genetic algorithm with multiple crossover operations to select variables from the expanded basis. Using new product launch data from the biochemical industry, we found that the proposed algorithm offers mean percentage decreases in the misclassification error rate of 50%, 56%, 59%, 77%, and 78% in comparison to a support vector machine, artificial neural network, quadratic discriminant analysis, linear discriminant analysis, and logistic regression, respectively. These improvements correspond to annual cost savings of $4.40–$25.73 million.     

Identification of Influential Cases in Kernel Fisher Discriminant Analysis

We study the influence of a single data case on the results of a statistical analysis. This problem has been addressed in several articles for linear discriminant analysis (LDA). Kernel Fisher discriminant analysis (KFDA) is a kernel based extension of LDA. In this article, we study the effect of atypical data points on KFDA and develop criteria for identification of cases having a detrimental effect on the classification performance of the KFDA classifier. We find that the criteria are successful in identifying cases whose omission from the training data prior to obtaining the KFDA classifier results in reduced error rates.     

Comparisons of decision tree methods using water data

This article demonstrates the application of classification trees (decision trees), logistic regression (LR), and linear discriminant function (LDR) to classify data of water quality (i.e., whether the water is fit for drinking on not fit for drinking). The data on water quality were obtained from Pakistan Council of Research in Water Resources (PCRWR) for two cities of Pakistan—one representing industrial environment (Sialkot) and the other one representing non-industrial environment (Narowal). To classify data on water quality, three statistical tools were employed—the Decision Tree methodology using Gini Index, LR, and LDA—using R software library. The results obtained by the said three techniques were compared using misclassification rates (a model with minimum value of misclassification rate is better). It was witnessed that LR performed well than the other two techniques while the Decision trees and LDA performed equally well. But for illustration purposes decision trees technique is comparatively easy to draw and interpret.     

Robust Model-Free Multiclass Probability Estimation

Classical statistical approaches for multiclass probability estimation are typically based on regression techniques such as multiple logistic regression, or density estimation approaches such as linear discriminant analysis (LDA) and quadratic discriminant analysis (QDA). These methods often make certain assumptions on the form of probability functions or on the underlying distributions of subclasses. In this article, we develop a model-free procedure to estimate multiclass probabilities based on large-margin classifiers. In particular, the new estimation scheme is employed by solving a series of weighted large-margin classifiers and then systematically extracting the probability information from these multiple classification rules. A main advantage of the proposed probability estimation technique is that it does not impose any strong parametric assumption on the underlying distribution and can be applied for a wide range of large-margin classification methods. A general computational algorithm is developed for class probability estimation. Furthermore, we establish asymptotic consistency of the probability estimates. Both simulated and real data examples are presented to illustrate competitive performance of the new approach and compare it with several other existing methods.     

Classification of biomedical signals for differential diagnosis of Raynaud's phenomenon

This paper discusses a supervised classification approach for the differential diagnosis of Raynaud's phenomenon (RP). The classification of data from healthy subjects and from patients suffering for primary and secondary RP is obtained by means of a set of classifiers derived within the framework of linear discriminant analysis. A set of functional variables and shape measures extracted from rewarming/reperfusion curves are proposed as discriminant features. Since the prediction of group membership is based on a large number of these features, the high dimension/small sample size problem is considered to overcome the singularity problem of the within-group covariance matrix. Results on a data set of 72 subjects demonstrate that a satisfactory classification of the subjects can be achieved through the proposed methodology.     

Modified linear discriminant analysis using block covariance matrix in high-dimensional data

Among many classification methods, linear discriminant analysis (LDA) is a favored tool due to its simplicity, robustness, and predictive accuracy but when the number of genes is larger than the number of observations, it cannot be applied directly because the within-class covariance matrix is singular. Also, diagonal LDA (DLDA) is a simpler model compared to LDA and has better performance in some cases. However, in reality, DLDA requires a strong assumption based on mutual independence. In this article, we propose the modified LDA (MLDA). MLDA is based on independence, but uses the information that has an effect on classification performance with the dependence structure. We suggest two approaches. One is the case of using gene rank. The other involves no use of gene rank. We found that MLDA has better performance than LDA, DLDA, or K-nearest neighborhood and is comparable with support vector machines in real data analysis and the simulation study.     

Variable Selection for Kernel Classification

In this article, a variable selection procedure, called surrogate selection, is proposed which can be applied when a support vector machine or kernel Fisher discriminant analysis is used in a binary classification problem. Surrogate selection applies the lasso after substituting the kernel discriminant scores for the binary group labels, as well as values for the input variable observations. Empirical results are reported, showing that surrogate selection performs well.     

Penalized classification using Fisher's linear discriminant

We consider the supervised classification setting, in which the data consist of p features measured on n observations, each of which belongs to one of K classes. Linear discriminant analysis (LDA) is a classical method for this problem. However, in the high-dimensional setting where p ? n, LDA is not appropriate for two reasons. First, the standard estimate for the within-class covariance matrix is singular, and so the usual discriminant rule cannot be applied. Second, when p is large, it is difficult to interpret the classification rule obtained from LDA, since it involves all p features. We propose penalized LDA, a general approach for penalizing the discriminant vectors in Fisher's discriminant problem in a way that leads to greater interpretability. The discriminant problem is not convex, so we use a minorization-maximization approach in order to efficiently optimize it when convex penalties are applied to the discriminant vectors. In particular, we consider the use of L(1) and fused lasso penalties. Our proposal is equivalent to recasting Fisher's discriminant problem as a biconvex problem. We evaluate the performances of the resulting methods on a simulation study, and on three gene expression data sets. We also survey past methods for extending LDA to the high-dimensional setting, and explore their relationships with our proposal.     

Functional Analysis of Variance,Discriminant Analysis,and Clustering in a Manifold of Elastic Curves

We develop functional data analysis techniques using the differential geometry of a manifold of smooth elastic functions on an interval in which the functions are represented by a log-speed function and an angle function. The manifold's geometry provides a method for computing a sample mean function and principal components on tangent spaces. Using tangent principal component analysis, we estimate probability models for functional data and apply them to functional analysis of variance, discriminant analysis, and clustering. We demonstrate these tasks using a collection of growth curves from children from ages 1–18.     

Robust selection of variables in linear discriminant analysis

A commonly used procedure for reduction of the number of variables in linear discriminant analysis is the stepwise method for variable selection. Although often criticized, when used carefully, this method can be a useful prelude to a further analysis. The contribution of a variable to the discriminatory power of the model is usually measured by the maximum likelihood ratio criterion, referred to as Wilks’ lambda. It is well known that the Wilks’ lambda statistic is extremely sensitive to the influence of outliers. In this work a robust version of the Wilks’ lambda statistic will be constructed based on the Minimum Covariance Discriminant (MCD) estimator and its reweighed version which has a higher efficiency. Taking advantage of the availability of a fast algorithm for computing the MCD a simulation study will be done to evaluate the performance of this statistic. The presentation of material in this article does not imply the expression of any opinion whatsoever on the part of Austro Control GmbH and is the sole responsibility of the authors.     

Sparse subspace linear discriminant analysis

We study high dimensional multigroup classification from a sparse subspace estimation perspective, unifying the linear discriminant analysis (LDA) with other recent developments in high dimensional multivariate analysis using similar tools, such as penalization method. We develop two two-stage sparse LDA models, where in the first stage, convex relaxation is used to convert two classical formulations of LDA to semidefinite programs, and furthermore subspace perspective allows for straightforward regularization and estimation. After the initial convex relaxation, we use a refinement stage to improve the accuracy. For the first model, a penalized quadratic program with group lasso penalty is used for refinement, whereas a sparse version of the power method is used for the second model. We carefully examine the theoretical properties of both methods, alongside with simulations and real data analysis.