import numpy as np import music_utils from one_vs_all import one_vs_allLogisticRegressor from sklearn import cross_validation from sklearn.metrics import confusion_matrix, classification_report # some global constants MUSIC_DIR = "music/" genres = ["blues","classical","country","disco","hiphop","jazz","metal","pop","reggae","rock"] # select the CEPS or FFT representation X,y = music_utils.read_ceps(genres,MUSIC_DIR) Xfft,yfft = music_utils.read_fft(genres,MUSIC_DIR) # select a regularization parameter reg = 1.0 # create a 1-vs-all classifier ova_logreg = one_vs_allLogisticRegressor(np.arange(10)) ova_logreg_fft = one_vs_allLogisticRegressor(np.arange(10)) # divide X into train and test sets X_train, X_test, y_train, y_test = cross_validation.train_test_split(X, y, test_size=0.2) Xfft_train, Xfft_test, yfft_train, yfft_test = cross_validation.train_test_split(Xfft, yfft, test_size=0.2) # train the K classifiers in 1-vs-all mode
import numpy as np import music_utils from one_vs_all import one_vs_allLogisticRegressor from sklearn import cross_validation from sklearn.metrics import confusion_matrix, classification_report, accuracy_score # some global constants MUSIC_DIR = "music/" genres = ["blues","classical","country","disco","hiphop","jazz","metal","pop","reggae","rock"] # select the CEPS or FFT representation # X,y = music_utils.read_ceps(genres,MUSIC_DIR) X,y = music_utils.read_fft(genres,MUSIC_DIR) # select a regularization parameter reg = 50.0 # create a 1-vs-all classifier ova_logreg = one_vs_allLogisticRegressor(np.arange(10)) # divide X into train and test sets X_train, X_test, y_train, y_test = cross_validation.train_test_split(X, y, test_size=0.2) # train the K classifiers in 1-vs-all mode ova_logreg.train(X_train,y_train,reg,'l1')