Showing papers on "Feature selection published in 2009"

Normalized Mutual Information Feature Selection

Abstract: A filter method of feature selection based on mutual information, called normalized mutual information feature selection (NMIFS), is presented. NMIFS is an enhancement over Battiti's MIFS, MIFS-U, and mRMR methods. The average normalized mutual information is proposed as a measure of redundancy among features. NMIFS outperformed MIFS, MIFS-U, and mRMR on several artificial and benchmark data sets without requiring a user-defined parameter. In addition, NMIFS is combined with a genetic algorithm to form a hybrid filter/wrapper method called GAMIFS. This includes an initialization procedure and a mutation operator based on NMIFS to speed up the convergence of the genetic algorithm. GAMIFS overcomes the limitations of incremental search algorithms that are unable to find dependencies between groups of features.

A review of Bayesian variable selection methods: what, how and which

Abstract: The selection of variables in regression problems has occupied the minds of many statisticians. Several Bayesian variable selection methods have been developed, and we concentrate on the following methods: Kuo & Mallick, Gibbs Variable Selection (GVS), Stochastic Search Variable Selection (SSVS), adaptive shrinkage with Jeffreys' prior or a Laplacian prior, and reversible jump MCMC. We review these methods, in the context of their different properties. We then implement the methods in BUGS, using both real and simulated data as examples, and investigate how the different methods perform in practice. Our results suggest that SSVS, reversible jump MCMC and adaptive shrinkage methods can all work well, but the choice of which method is better will depend on the priors that are used, and also on how they are implemented.

A comparison of random forest and its Gini importance with standard chemometric methods for the feature selection and classification of spectral data

Abstract: Regularized regression methods such as principal component or partial least squares regression perform well in learning tasks on high dimensional spectral data, but cannot explicitly eliminate irrelevant features. The random forest classifier with its associated Gini feature importance, on the other hand, allows for an explicit feature elimination, but may not be optimally adapted to spectral data due to the topology of its constituent classification trees which are based on orthogonal splits in feature space. We propose to combine the best of both approaches, and evaluated the joint use of a feature selection based on a recursive feature elimination using the Gini importance of random forests' together with regularized classification methods on spectral data sets from medical diagnostics, chemotaxonomy, biomedical analytics, food science, and synthetically modified spectral data. Here, a feature selection using the Gini feature importance with a regularized classification by discriminant partial least squares regression performed as well as or better than a filtering according to different univariate statistical tests, or using regression coefficients in a backward feature elimination. It outperformed the direct application of the random forest classifier, or the direct application of the regularized classifiers on the full set of features. The Gini importance of the random forest provided superior means for measuring feature relevance on spectral data, but – on an optimal subset of features – the regularized classifiers might be preferable over the random forest classifier, in spite of their limitation to model linear dependencies only. A feature selection based on Gini importance, however, may precede a regularized linear classification to identify this optimal subset of features, and to earn a double benefit of both dimensionality reduction and the elimination of noise from the classification task.

L1 Penalized Estimation in the Cox Proportional Hazards Model

Abstract: This article presents a novel algorithm that efficiently computes L(1) penalized (lasso) estimates of parameters in high-dimensional models. The lasso has the property that it simultaneously performs variable selection and shrinkage, which makes it very useful for finding interpretable prediction rules in high-dimensional data. The new algorithm is based on a combination of gradient ascent optimization with the Newton-Raphson algorithm. It is described for a general likelihood function and can be applied in generalized linear models and other models with an L(1) penalty. The algorithm is demonstrated in the Cox proportional hazards model, predicting survival of breast cancer patients using gene expression data, and its performance is compared with competing approaches. An R package, penalized, that implements the method, is available on CRAN.

Environmental Sound Recognition With Time–Frequency Audio Features

Abstract: The paper considers the task of recognizing environmental sounds for the understanding of a scene or context surrounding an audio sensor. A variety of features have been proposed for audio recognition, including the popular Mel-frequency cepstral coefficients (MFCCs) which describe the audio spectral shape. Environmental sounds, such as chirpings of insects and sounds of rain which are typically noise-like with a broad flat spectrum, may include strong temporal domain signatures. However, only few temporal-domain features have been developed to characterize such diverse audio signals previously. Here, we perform an empirical feature analysis for audio environment characterization and propose to use the matching pursuit (MP) algorithm to obtain effective time-frequency features. The MP-based method utilizes a dictionary of atoms for feature selection, resulting in a flexible, intuitive and physically interpretable set of features. The MP-based feature is adopted to supplement the MFCC features to yield higher recognition accuracy for environmental sounds. Extensive experiments are conducted to demonstrate the effectiveness of these joint features for unstructured environmental sound classification, including listening tests to study human recognition capabilities. Our recognition system has shown to produce comparable performance as human listeners.

The composite absolute penalties family for grouped and hierarchical variable selection

Abstract: Extracting useful information from high-dimensional data is an important focus of today's statistical research and practice. Penalized loss function minimization has been shown to be effective for this task both theoretically and empirically. With the virtues of both regularization and sparsity, the $L_1$-penalized squared error minimization method Lasso has been popular in regression models and beyond. In this paper, we combine different norms including $L_1$ to form an intelligent penalty in order to add side information to the fitting of a regression or classification model to obtain reasonable estimates. Specifically, we introduce the Composite Absolute Penalties (CAP) family, which allows given grouping and hierarchical relationships between the predictors to be expressed. CAP penalties are built by defining groups and combining the properties of norm penalties at the across-group and within-group levels. Grouped selection occurs for nonoverlapping groups. Hierarchical variable selection is reached by defining groups with particular overlapping patterns. We propose using the BLASSO and cross-validation to compute CAP estimates in general. For a subfamily of CAP estimates involving only the $L_1$ and $L_{\infty}$ norms, we introduce the iCAP algorithm to trace the entire regularization path for the grouped selection problem. Within this subfamily, unbiased estimates of the degrees of freedom (df) are derived so that the regularization parameter is selected without cross-validation. CAP is shown to improve on the predictive performance of the LASSO in a series of simulated experiments, including cases with $p\gg n$ and possibly mis-specified groupings. When the complexity of a model is properly calculated, iCAP is seen to be parsimonious in the experiments.

The composite absolute penalties family for grouped and hierarchical variable selection

Abstract: Extracting useful information from high-dimensional data is an important focus of today’s statistical research and practice. Penalized loss function minimization has been shown to be effective for this task both theoretically and empirically. With the virtues of both regularization and sparsity, the L1-penalized squared error minimization method Lasso has been popular in regression models and beyond. In this paper, we combine different norms including L1 to form an intelligent penalty in order to add side information to the fitting of a regression or classification model to obtain reasonable estimates. Specifically, we introduce the Composite Absolute Penalties (CAP) family, which allows given grouping and hierarchical relationships between the predictors to be expressed. CAP penalties are built by defining groups and combining the properties of norm penalties at the across-group and within-group levels. Grouped selection occurs for nonoverlapping groups. Hierarchical variable selection is reached by defining groups with particular overlapping patterns. We propose using the BLASSO and cross-validation to compute CAP estimates in general. For a subfamily of CAP estimates involving only the L1 and L∞ norms, we introduce the iCAP algorithm to trace the entire regularization path for the grouped selection problem. Within this subfamily, unbiased estimates of the degrees of freedom (df) are derived so that the regularization parameter is selected without cross-validation. CAP is shown to improve on the predictive performance of the LASSO in a series of simulated experiments, including cases with p≫n and possibly mis-specified groupings. When the complexity of a model is properly calculated, iCAP is seen to be parsimonious in the experiments.

High-dimensional variable selection

Abstract: This paper explores the following question: what kind of statistical guarantees can be given when doing variable selection in high dimensional models? In particular, we look at the error rates and power of some multi-stage regression methods. In the first stage we fit a set of candidate models. In the second stage we select one model by cross-validation. In the third stage we use hypothesis testing to eliminate some variables. We refer to the first two stages as "screening" and the last stage as "cleaning." We consider three screening methods: the lasso, marginal regression, and forward stepwise regression. Our method gives consistent variable selection under certain conditions.

Feature selection for text classification with Naïve Bayes

Abstract: As an important preprocessing technology in text classification, feature selection can improve the scalability, efficiency and accuracy of a text classifier. In general, a good feature selection method should consider domain and algorithm characteristics. As the Naive Bayesian classifier is very simple and efficient and highly sensitive to feature selection, so the research of feature selection specially for it is significant. This paper presents two feature evaluation metrics for the Naive Bayesian classifier applied on multi-class text datasets: Multi-class Odds Ratio (MOR), and Class Discriminating Measure (CDM). Experiments of text classification with Naive Bayesian classifiers were carried out on two multi-class texts collections. As the results indicate, CDM and MOR gain obviously better selecting effect than other feature selection approaches.

New Approaches to Fuzzy-Rough Feature Selection

Abstract: There has been great interest in developing methodologies that are capable of dealing with imprecision and uncertainty. The large amount of research currently being carried out in fuzzy and rough sets is representative of this. Many deep relationships have been established, and recent studies have concluded as to the complementary nature of the two methodologies. Therefore, it is desirable to extend and hybridize the underlying concepts to deal with additional aspects of data imperfection. Such developments offer a high degree of flexibility and provide robust solutions and advanced tools for data analysis. Fuzzy-rough set-based feature (FS) selection has been shown to be highly useful at reducing data dimensionality but possesses several problems that render it ineffective for large datasets. This paper proposes three new approaches to fuzzy-rough FS-based on fuzzy similarity relations. In particular, a fuzzy extension to crisp discernibility matrices is proposed and utilized. Initial experimentation shows that the methods greatly reduce dimensionality while preserving classification accuracy.

Multi-task feature learning via efficient l 2, 1 -norm minimization

Abstract: The problem of joint feature selection across a group of related tasks has applications in many areas including biomedical informatics and computer vision. We consider the l2, 1-norm regularized regression model for joint feature selection from multiple tasks, which can be derived in the probabilistic framework by assuming a suitable prior from the exponential family. One appealing feature of the l2, 1-norm regularization is that it encourages multiple predictors to share similar sparsity patterns. However, the resulting optimization problem is challenging to solve due to the non-smoothness of the l2, 1-norm regularization. In this paper, we propose to accelerate the computation by reformulating it as two equivalent smooth convex optimization problems which are then solved via the Nesterov's method---an optimal first-order black-box method for smooth convex optimization. A key building block in solving the reformulations is the Euclidean projection. We show that the Euclidean projection for the first reformulation can be analytically computed, while the Euclidean projection for the second one can be computed in linear time. Empirical evaluations on several data sets verify the efficiency of the proposed algorithms.

Robust biomarker identification for cancer diagnosis using ensemble feature selection methods

Abstract: Motivation: Biomarker discovery is an important topic in biomedical applications of computational biology, including applications such as gene and SNP selection from high dimensional data. Surprisingly, the stability with respect to sampling variation or robustness of such selection processes has received attention only recently. However, robustness of biomarkers is an important issue, as it may greatly influence subsequent biological validations. In addition, a more robust set of markers may strengthen the confidence of an expert in the results of a selection method. Results: Our first contribution is a general framework for the analysis of the robustness of a biomarker selection algorithm. Secondly, we conducted a large-scale analysis of the recently introduced concept of ensemble feature selection, where multiple feature selections are combined in order to increase the robustness of the final set of selected features. We focus on selection methods that are embedded in the estimation of support vector machines (SVMs). SVMs are powerful classification models that have shown state-ofthe-art performance on several diagnosis and prognosis tasks on biological data. Their feature selection extensions also offered good results for gene selection tasks. We show that the robustness of SVMs for biomarker discovery can be substantially increased by using ensemble feature selection techniques, while at the same time improving upon classification performances. The proposed methodology is evaluated on four microarray data sets showing increases of up to almost 30% in robustness of the selected biomarkers, along with an improvement of about 15% in classification performance. The stability improvement with ensemble methods is particularly noticeable for small signature sizes (a few tens of genes), which is most relevant for the design of a diagnosis or prognosis model from a gene signature.

More generality in efficient multiple kernel learning

Abstract: Recent advances in Multiple Kernel Learning (MKL) have positioned it as an attractive tool for tackling many supervised learning tasks. The development of efficient gradient descent based optimization schemes has made it possible to tackle large scale problems. Simultaneously, MKL based algorithms have achieved very good results on challenging real world applications. Yet, despite their successes, MKL approaches are limited in that they focus on learning a linear combination of given base kernels.In this paper, we observe that existing MKL formulations can be extended to learn general kernel combinations subject to general regularization. This can be achieved while retaining all the efficiency of existing large scale optimization algorithms. To highlight the advantages of generalized kernel learning, we tackle feature selection problems on benchmark vision and UCI databases. It is demonstrated that the proposed formulation can lead to better results not only as compared to traditional MKL but also as compared to state-of-the-art wrapper and filter methods for feature selection.

Ultrahigh Dimensional Feature Selection: Beyond The Linear Model

Abstract: Variable selection in high-dimensional space characterizes many contemporary problems in scientific discovery and decision making. Many frequently-used techniques are based on independence screening; examples include correlation ranking (Fan & Lv, 2008) or feature selection using a two-sample t-test in high-dimensional classification (Tibshirani et al., 2003). Within the context of the linear model, Fan & Lv (2008) showed that this simple correlation ranking possesses a sure independence screening property under certain conditions and that its revision, called iteratively sure independent screening (ISIS), is needed when the features are marginally unrelated but jointly related to the response variable. In this paper, we extend ISIS, without explicit definition of residuals, to a general pseudo-likelihood framework, which includes generalized linear models as a special case. Even in the least-squares setting, the new method improves ISIS by allowing feature deletion in the iterative process. Our technique allows us to select important features in high-dimensional classification where the popularly used two-sample t-method fails. A new technique is introduced to reduce the false selection rate in the feature screening stage. Several simulated and two real data examples are presented to illustrate the methodology.

Shrinkage tuning parameter selection with a diverging number of parameters

Abstract: Contemporary statistical research frequently deals with problems involving a diverging number of parameters. For those problems, various shrinkage methods (e.g. the lasso and smoothly clipped absolute deviation) are found to be particularly useful for variable selection. Nevertheless, the desirable performances of those shrinkage methods heavily hinge on an appropriate selection of the tuning parameters. With a fixed predictor dimension, Wang and co-worker have demonstrated that the tuning parameters selected by a Bayesian information criterion type criterion can identify the true model consistently. In this work, similar results are further extended to the situation with a diverging number of parameters for both unpenalized and penalized estimators. Consequently, our theoretical results further enlarge not only the scope of applicability of the traditional Bayesian information criterion type criteria but also that of those shrinkage estimation methods.

Disease state prediction from resting state functional connectivity

Abstract: The application of multivoxel pattern analysis methods has attracted increasing attention, particularly for brain state prediction and real-time functional MRI applications. Support vector classification is the most popular of these techniques, owing to reports that it has better prediction accuracy and is less sensitive to noise. Support vector classification was applied to learn functional connectivity patterns that distinguish patients with depression from healthy volunteers. In addition, two feature selection algorithms were implemented (one filter method, one wrapper method) that incorporate reliability information into the feature selection process. These reliability feature selections methods were compared to two previously proposed feature selection methods. A support vector classifier was trained that reliably distinguishes healthy volunteers from clinically depressed patients. The reliability feature selection methods outperformed previously utilized methods. The proposed framework for applying support vector classification to functional connectivity data is applicable to other disease states beyond major depression.

Recognising action as clouds of space-time interest points

Abstract: Much of recent action recognition research is based on space-time interest points extracted from video using a Bag of Words (BOW) representation. It mainly relies on the discriminative power of individual local space-time descriptors, whilst ignoring potentially valuable information about the global spatio-temporal distribution of interest points. In this paper, we propose a novel action recognition approach which differs significantly from previous interest points based approaches in that only the global spatiotemporal distribution of the interest points are exploited. This is achieved through extracting holistic features from clouds of interest points accumulated over multiple temporal scales followed by automatic feature selection. Our approach avoids the non-trivial problems of selecting the optimal space-time descriptor, clustering algorithm for constructing a codebook, and selecting codebook size faced by previous interest points based methods. Our model is able to capture smooth motions, robust to view changes and occlusions at a low computation cost. Experiments using the KTH and WEIZMANN datasets demonstrate that our approach outperforms most existing methods.

Text feature selection using ant colony optimization

Abstract: Feature selection and feature extraction are the most important steps in classification systems. Feature selection is commonly used to reduce dimensionality of datasets with tens or hundreds of thousands of features which would be impossible to process further. One of the problems in which feature selection is essential is text categorization. A major problem of text categorization is the high dimensionality of the feature space; therefore, feature selection is the most important step in text categorization. At present there are many methods to deal with text feature selection. To improve the performance of text categorization, we present a novel feature selection algorithm that is based on ant colony optimization. Ant colony optimization algorithm is inspired by observation on real ants in their search for the shortest paths to food sources. Proposed algorithm is easily implemented and because of use of a simple classifier in that, its computational complexity is very low. The performance of proposed algorithm is compared to the performance of genetic algorithm, information gain and CHI on the task of feature selection in Reuters-21578 dataset. Simulation results on Reuters-21578 dataset show the superiority of the proposed algorithm.

Sparse Canonical Correlation Analysis with Application to Genomic Data Integration

Abstract: Large scale genomic studies with multiple phenotypic or genotypic measures may require the identification of complex multivariate relationships. In multivariate analysis a common way to inspect the relationship between two sets of variables based on their correlation is canonical correlation analysis, which determines linear combinations of all variables of each type with maximal correlation between the two linear combinations. However, in high dimensional data analysis, when the number of variables under consideration exceeds tens of thousands, linear combinations of the entire sets of features may lack biological plausibility and interpretability. In addition, insufficient sample size may lead to computational problems, inaccurate estimates of parameters and non-generalizable results. These problems may be solved by selecting sparse subsets of variables, i.e. obtaining sparse loadings in the linear combinations of variables of each type. In this paper we present Sparse Canonical Correlation Analysis (SCCA) which examines the relationships between two types of variables and provides sparse solutions that include only small subsets of variables of each type by maximizing the correlation between the subsets of variables of different types while performing variable selection. We also present an extension of SCCA--adaptive SCCA. We evaluate their properties using simulated data and illustrate practical use by applying both methods to the study of natural variation in human gene expression.

On the Use of the Adjusted Rand Index as a Metric for Evaluating Supervised Classification

Abstract: The Adjusted Rand Index (ARI) is frequently used in cluster validation since it is a measure of agreement between two partitions: one given by the clustering process and the other defined by external criteria. In this paper we investigate the usability of this clustering validation measure in supervised classification problems by two different approaches: as a performance measure and in feature selection. Since ARI measures the relation between pairs of dataset elements not using information from classes (labels) it can be used to detect problems with the classification algorithm specially when combined with conventional performance measures. Instead, if we use the class information, we can apply ARI also to perform feature selection. We present the results of several experiments where we have applied ARI both as a performance measure and for feature selection showing the validity of this index for the given tasks.

Performance of several variable-selection methods applied to real ecological data.

Abstract: I evaluated the predictive ability of statistical models obtained by applying seven methods of variable selection to 12 ecological and environmental data sets. Cross-validation, involving repeated splits of each data set into training and validation subsets, was used to obtain honest estimates of predictive ability that could be fairly compared among methods. There was surprisingly little difference in predictive ability among five methods based on multiple linear regression. Stepwise methods performed similarly to exhaustive algorithms for subset selection, and the choice of criterion for comparing models (Akaike's information criterion, Schwarz's Bayesian information criterion or F statistics) had little effect on predictive ability. For most of the data sets, two methods based on regression trees yielded models with substantially lower predictive ability. I argue that there is no 'best' method of variable selection and that any of the regression-based approaches discussed here is capable of yielding useful predictive models.

A group bridge approach for variable selection

Abstract: In multiple regression problems when covariates can be naturally grouped, it is important to carry out feature selection at the group and within-group individual variable levels simultaneously. The existing methods, including the lasso and group lasso, are designed for either variable selection or group selection, but not for both. We propose a group bridge approach that is capable of simultaneous selection at both the group and within-group individual variable levels. The proposed approach is a penalized regularization method that uses a specially designed group bridge penalty. It has the oracle group selection property, in that it can correctly select important groups with probability converging to one. In contrast, the group lasso and group least angle regression methods in general do not possess such an oracle property in group selection. Simulation studies indicate that the group bridge has superior performance in group and individual variable selection relative to several existing methods.

Steganalysis by subtractive pixel adjacency matrix

Abstract: This paper presents a novel method for detection of steganographic methods that embed in the spatial domain by adding a low-amplitude independent stego signal, an example of which is LSB matching. First, arguments are provided for modeling differences between adjacent pixels using first-order and second-order Markov chains. Subsets of sample transition probability matrices are then used as features for a steganalyzer implemented by support vector machines. The accuracy of the presented steganalyzer is evaluated on LSB matching and four different databases. The steganalyzer achieves superior accuracy with respect to prior art and provides stable results across various cover sources. Since the feature set based on second-order Markov chain is high-dimensional, we address the issue of curse of dimensionality using a feature selection algorithm and show that the curse did not occur in our experiments.

Discriminant Saliency, the Detection of Suspicious Coincidences, and Applications to Visual Recognition

Abstract: A discriminant formulation of top-down visual saliency, intrinsically connected to the recognition problem, is proposed. The new formulation is shown to be closely related to a number of classical principles for the organization of perceptual systems, including infomax, inference by detection of suspicious coincidences, classification with minimal uncertainty, and classification with minimum probability of error. The implementation of these principles with computational parsimony, by exploitation of the statistics of natural images, is investigated. It is shown that Barlow's principle of inference by the detection of suspicious coincidences enables computationally efficient saliency measures which are nearly optimal for classification. This principle is adopted for the solution of the two fundamental problems in discriminant saliency, feature selection and saliency detection. The resulting saliency detector is shown to have a number of interesting properties, and act effectively as a focus of attention mechanism for the selection of interest points according to their relevance for visual recognition. Experimental evidence shows that the selected points have good performance with respect to 1) the ability to localize objects embedded in significant amounts of clutter, 2) the ability to capture information relevant for image classification, and 3) the richness of the set of visual attributes that can be considered salient.