De-Codec: Decoupling Speech and Background-Sound in Codec to Allow Explicit Feature Selections for Downstream Audio Tasks

Implicit feature identification in Chinese reviews using explicit topic mining model

Primal explicit max margin feature selection for nonlinear support vector machines

Explicit unsupervised feature selection based on structured graph and locally linear embedding

With the increasing number of mobile Apps developed, they are now closely integrated into daily life. In this paper, we develop a framework to predict mobile Apps that are most likely to be used regarding the current device status of a smartphone. Such an Apps usage prediction framework is a crucial prerequisite for fast App launching, intelligent user experience, and power management of smartphones. By analyzing real App usage log data, we discover two kinds of features: The Explicit Feature (EF) from sensing readings of built-in sensors, and the Implicit Feature (IF) from App usage relations. The IF feature is derived by constructing the proposed App Usage Graph (abbreviated as AUG) that models App usage transitions. In light of AUG, we are able to discover usage relations among Apps. Since users may have different usage behaviors on their smartphones, we further propose one personalized feature selection algorithm. We explore minimum description length (MDL) from the training data and select those features which need less length to describe the training data. The personalized feature selection can successfully reduce the log size and the prediction time. Finally, we adopt the kNN classification model to predict Apps usage. Note that through the features selected by the proposed personalized feature selection algorithm, we only need to keep these features, which in turn reduces the prediction time and avoids the curse of dimensionality when using the kNN classifier. The results based on a real dataset demonstrate the effectiveness of the proposed framework and show the predictive capability for App usage prediction.

The use of compression algorithms in machine learning tasks such as clustering and classification has appeared in a variety of fields, sometimes with the promise of reducing problems of explicit feature selection. The theoretical justification for such methods has been founded on an upper bound on Kolmogorov complexity and an idealized information space. An alternate view shows compression algorithms implicitly map strings into implicit feature space vectors, and compression-based similarity measures compute similarity within these feature spaces. Thus, compression-based methods are not a parameter free magic bullet for feature selection and data representation, but are instead concrete similarity measures within defined feature spaces, and are therefore akin to explicit feature vector models used in standard machine learning algorithms. To underscore this point, we find theoretical and empirical connections between traditional machine learning vector models and compression, encouraging cross-fertilization in future work.

In this paper, we present a new adaptive feature scaling scheme for ultrahigh-dimensional feature selection on Big Data, and then reformulate it as a convex semi-infinite programming (SIP) problem. To address the SIP, we propose an efficient feature generating paradigm. Different from traditional gradient-based approaches that conduct optimization on all input features, the proposed paradigm iteratively activates a group of features, and solves a sequence of multiple kernel learning (MKL) subproblems. To further speed up the training, we propose to solve the MKL subproblems in their primal forms through a modified accelerated proximal gradient approach. Due to such optimization scheme, some efficient cache techniques are also developed. The feature generating paradigm is guaranteed to converge globally under mild conditions, and can achieve lower feature selection bias. Moreover, the proposed method can tackle two challenging tasks in feature selection: 1) group-based feature selection with complex structures, and 2) nonlinear feature selection with explicit feature mappings. Comprehensive experiments on a wide range of synthetic and real-world data sets of tens of million data points with O(1014) features demonstrate the competitive performance of the proposed method over state-of-the-art feature selection methods in terms of generalization performance and training effciency.

In this paper, we present an audio segmentation system for broadcast news, and its results in the Albayzin-2010 evaluation. First of all, the Albayzin-2010 evaluation setup, developed by the authors, is presented; in particular, the database and the metric are described. The reported hierarchical HMM-GMM-based system is composed of one binary detector for each of the five considered classes (music, speech, speech over music, speech over noise and other). A fast one-pass-training feature selection technique is adapted to the audio segmentation task to improve the results and to reduce the dimensionality of the input feature vector.

Nonlinear classification problems are always assumed to be equivalent to a linear classification problem in some higher dimensional feature space. Kernel machines like support vector machines (SVMs) implicitly map the features to higher dimensional feature space by a kernel trick, and use all the mapped features for classification. This paper proposes an explicit polynomial feature expansion and feature selection method, which eliminates 'useless' features before designing a linear classifier. This method is almost automatic, and it is easier to use than selecting kernel functions and parameters for kernel machines. In our experiments, the method achieves good generalization ability on most of the benchmark datasets, and it can be a candidate method for solving general pattern recognition problems.

Feature matching, which refers to establishing the correspondence of regions between two images (usually voxel features), is a crucial prerequisite of feature-based registration. For deformable image registration tasks, traditional feature-based registration methods typically use an iterative matching strategy for interest region matching, where feature selection and matching are explicit, but specific feature selection schemes are often useful in solving application-specific problems and require several minutes for each registration. In the past few years, the feasibility of learning-based methods, such as VoxelMorph and TransMorph, has been proven, and their performance has been shown to be competitive compared to traditional methods. However, these methods are usually single-stream, where the two images to be registered are concatenated into a 2-channel whole, and then the deformation field is output directly. The transformation of image features into interimage matching relationships is implicit. In this paper, we propose a novel end-to-end dual-stream unsupervised framework, named TransMatch, where each image is fed into a separate stream branch, and each branch performs feature extraction independently. Then, we implement explicit multilevel feature matching between image pairs via the query-key matching idea of the self-attention mechanism in the Transformer model. Comprehensive experiments are conducted on three 3D brain MR datasets, LPBA40, IXI, and OASIS, and the results show that the proposed method achieves state-of-the-art performance in several evaluation metrics compared to the commonly utilized registration methods, including SyN, NiftyReg, VoxelMorph, CycleMorph, ViT-V-Net, and TransMorph, demonstrating the effectiveness of our model in deformable medical image registration.

Feature selection for classification in high-dimensional spaces can improve generalization, reduce classifier complexity, and identify important, discriminating feature "markers." For support vector machine (SVM) classification, a widely used technique is recursive feature elimination (RFE). We demonstrate that RFE is not consistent with margin maximization, central to the SVM learning approach. We thus propose explicit margin-based feature elimination (MFE) for SVMs and demonstrate both improved margin and improved generalization, compared with RFE. Moreover, for the case of a nonlinear kernel, we show that RFE assumes that the squared weight vector 2-norm is strictly decreasing as features are eliminated. We demonstrate this is not true for the Gaussian kernel and, consequently, RFE may give poor results in this case. MFE for nonlinear kernels gives better margin and generalization. We also present an extension which achieves further margin gains, by optimizing only two degrees of freedom--the hyperplane's intercept and its squared 2-norm--with the weight vector orientation fixed. We finally introduce an extension that allows margin slackness. We compare against several alternatives, including RFE and a linear programming method that embeds feature selection within the classifier design. On high-dimensional gene microarray data sets, University of California at Irvine (UCI) repository data sets, and Alzheimer's disease brain image data, MFE methods give promising results.

On Expressive Features for Gait Analysis using Lower Limb Inertial Sensor Data

High-dimensional pattern classification problems with a small number of training patterns are difficult. This paper deals with classification of motor imagery tasks for brain-computer interfacing (BCI), which is a hard problem involving a relatively small number of high-dimensional training patterns where curse of dimensionality issue has to be taken into account and feature selection is an important requirement to build a suitable classifier. Evolutionary metaheuristics for feature selection are usually more time-consuming than other alternatives, but their high performances in terms of classification accuracy make them desirable approaches. In this paper, feature selection through a wrapper procedure based on multi-objective optimization is compared with the use of deep belief networks (DBN) that constitute powerful classifiers implementing feature selection implicitly. Two different classifiers, LDA (linear discriminant analysis) and DBN, have been used to classify EEG signals with features extracted by multiresolution analysis (MRA) and selected by a multiobjective evolutionary method that also uses LDA to implement the fitness function of the solutions. The experimental results show that DBNs usually provide better or similar classification performances without requiring an explicit feature selection phase. Nevertheless, the DBN’s classification performance significantly decreases in problems with a very large number of features. Moreover, to achieve high classification rates, it is necessary to determine a suitable structure for the DBN. Therefore, in this paper we also propose a multiobjective approach to tackle this problem.

Framework for efficient feature selection in genetic algorithm based data mining

Feature engineering is the task of improving predictive modelling performance on a dataset by transforming its feature space. Existing approaches to automate this process rely on either transformed feature space exploration through evaluation-guided search, or explicit expansion of datasets with all transformed features followed by feature selection. Such approaches incur high computational costs in runtime and/or memory. We present a novel technique, called Learning Feature Engineering (LFE), for automating feature engineering in classification tasks. LFE is based on learning the effectiveness of applying a transformation (e.g., arithmetic or aggregate operators) on numerical features, from past feature engineering experiences. Given a new dataset, LFE recommends a set of useful transformations to be applied on features without relying on model evaluation or explicit feature expansion and selection. Using a collection of datasets, we train a set of neural networks, which aim at predicting the transformation that impacts classification performance positively. Our empirical results show that LFE outperforms other feature engineering approaches for an overwhelming majority (89%) of the datasets from various sources while incurring a substantially lower computational cost.

At a conceptual level, one can divide the task of concept learning into the subtask of selecting a proper subset of features to use in describing the concept, and learning a hypothesis based on these features. This directly leads to a modular design of the learning algorithm which allows flexible combinations of explicit feature selection methods with model induction algorithms and sometimes leads to powerful variants. Many recent works, however, tend to take a more general view of feature selection as part of model selection and therefore integrate feature selection more closely into the learning algorithms (i.e. the Bayesian feature selection methods). Feature selection for clustering is a largely untouched problem, and there has been little theoretical characterization of the heuristic approaches we described in the chapter. In summary, although no universal strategy can be prescribed, for high-dimensional problems frequently encountered in microarray analysis, feature selection offers a promising suite of techniques to improve interpretability, performance and computation efficiency in learning.