deap_learning.rar

Feature Extraction: Firstly, extract features or load final_workspace.mat which contains a backup To extract features: 1) Download DEAP dataset and store it in code folder 2) Run psd_hamming_overlap.m to extract psd features. These are stored in a variable 'feature_vector' 3) Run 'msce_features_matrix.m' to extract MSCE features. These are stored in a variable ' msce_features_matrix' (Note: msce_features_matrix.m requires MATLAB 2013a or plus and many hours on 8GB machine) Alternatively, load 'final_workspace.mat' which has the 'feature_vector' and 'msce_features_matrix' the variables Running SVM with PCA and RBF Kernel and MSCE Features: Install 'libsvm' and set path Run pca_svm.m The gamma, c values and number of PCA dimensions can be altered inside pca_svm.m It performs 10 trials by default and reports mean and standard deviation of accuracies on console Running Neural Network Classifier with PCA: Run pca_neural.m This uses neural_classifier.m, verify.m and threshold.m The number of number of hidden nodes can be altered inside neural_classifier.m The number of PCA dimensions and number of trials can be changed in pca_neural.m The mean and standard deviation of accuracy is reported on console Running Logistic Regression: Run pca_logistic.m This uses reg.m and check_reg.m The stopping threshold for logistic regression can be changed from inside reg.m The number of PCA dimensions and number of trials can be changed in pca_logistic.m The mean and standard deviation of accuracy is reported on console Running SVM with PCA and MSCE Features on per person basis: Run pca_svm_personwise.m Number of persons, PCA dimensions, gamma and c values can be changed inside pca_svm_personwise.m Reports a vector of mean of accuracies for each person Reports a vector of std. of accuracies for each person Running Neural Network with PCA and MSCE Features on per person basis: Run pca_neural_personwise.m The number of number of hidden nodes can be altered inside neural_classifier.m Number of persons, PCA dimensions can be changed inside pca_svm_personwise.m Reports a vector of mean of accuracies for each person Reports a vector of std. of accuracies for each person Running Logistic Regression with PCA and MSCE Features on per person basis: Run pca_logistic_personwise.m This uses reg.m and check_reg.m This uses neural_classifier.m, verify.m and threshold.m The stopping threshold for logistic regression can be changed from inside reg.m Number of persons, PCA dimensions can be changed inside pca_svm_personwise.m Reports a vector of mean of accuracies for each person Reports a vector of std. of accuracies for each person