This paper investigates dimensionality reduction problem for signal decoding. Its main application is brain-computer interface modeling. The challenge is high redundancy in the data description. Data combines time ser...
详细信息
This paper investigates dimensionality reduction problem for signal decoding. Its main application is brain-computer interface modeling. The challenge is high redundancy in the data description. Data combines time series of two origins: design space: brain cortex signals and target space: limb motion signals. High correlations among measurements of complex signals lead to multiple correlations. This case studies correlations in both input and target spaces that carry heterogeneous data. This paper proposes featureselection algorithms to construct simple and stable forecasting model. It extends ideas of the quadratic programming feature selection approach and selects non-correlated features that are relevant to the target. The proposed methods take into account dependencies in both design and target space and select features, which fit both spaces jointly. The computational experiment was carried out using an electrocorticogram (ECoG) dataset. The obtained models predict hand motions using signals of the brain cortex. The partial least squares (PLS) regression model is used as the base model for dimensionality reduction. The best result is obtained by PLS algorithm, that reduces space dimensionality using the QPFS.
We reformulate the quadratic programming feature selection (QPFS) method in a Kernel space to obtain a vector which maximizes the quadratic objective function of QPFS. We demonstrate that the vector obtained by Kernel...
详细信息
We reformulate the quadratic programming feature selection (QPFS) method in a Kernel space to obtain a vector which maximizes the quadratic objective function of QPFS. We demonstrate that the vector obtained by Kernel quadratic programming feature selection is equivalent to the Kernel Fisher vector and, therefore, a new interpretation of the Kernel Fisher discriminant analysis is given which provides some computational advantages for highly unbalanced datasets. (C) 2011 Elsevier B.V. All rights reserved.
The paper is devoted to the problem of constructing a predictive model in the high-dimensional feature space. The space is redundant, there is multicollinearity in the design matrix columns. In this case the model is ...
详细信息
The paper is devoted to the problem of constructing a predictive model in the high-dimensional feature space. The space is redundant, there is multicollinearity in the design matrix columns. In this case the model is unstable to changes in data or in parameter values. To build a stable model, the authors solve the dimensionality reduction problem for the feature space. It is proposed to use featureselection methods during parameter optimization process. The idea is to select the active set of model parameters which have to be optimized in the current optimization step. quadratic programming feature selection is used to find the active set of parameters. The algorithm maximizes the relevance of model parameters to the residuals and makes them pairwise independent. Nonlinear regression and logistic regression models are investigated. We carried out the experiment to show how the proposed method works and compare it with other methods. The proposed algorithm achieves the less error and greater stability with comparison to the other methods.
In this study, we compared two feature extraction methods (PCA, PLS) and seven featureselection methods (mRMR and its variations, MaxRel, QPFS) on four different classifiers (SVM, RF, KNN, NN). We use ratio compariso...
详细信息
In this study, we compared two feature extraction methods (PCA, PLS) and seven featureselection methods (mRMR and its variations, MaxRel, QPFS) on four different classifiers (SVM, RF, KNN, NN). We use ratio comparison validation for PCA method and 10-folds cross validation method for both the feature extraction and featureselection methods. We use Leukemia data set and Colon data set to apply the combinations and measured accuracy as well as area under ROC. The results illustrated that featureselection and extraction methods can both somehow improve the performance of classification tasks on microarray data sets. Some combinations of classifier and feature preprocessing method can greatly improve the accuracy as well as the AUC value are given in this study.
暂无评论