def data():
# maxlen = 100
# max_features = 20000
frame_number= 50
emotions = ['sad','ang', 'neu', 'exc']
print('Loading data...')
X_train, y_train, weights = static_dataset('train','M',emotions,frame_number)
X_test, y_test = static_dataset('test','M',emotions,frame_number)
print(len(X_train), 'train sequences')
print(len(X_test), 'test sequences')
print("Pad sequences (samples x time)")
# X_train = sequence.pad_sequences(X_train, maxlen=maxlen)
# X_test = sequence.pad_sequences(X_test, maxlen=maxlen)
print('X_train shape:', X_train.shape)
print('X_test shape:', X_test.shape)
return X_train, X_test, y_train, y_test, weights
bias = True
drop_rate = 0.2
feature_number = 87
if feature_type == "LLF":
feature_number = 31
normal_FFNN = FFNN(trainable=trainable,
feature_number=feature_number,
frame_number=frame_number,
emotions=emotions,
lr=0.0001)
#MALE DATASET
x_tr_m, y_tr_m, class_weight_dict_m = static_dataset(feature_type, 'train',
'M', emotions,
frame_number)
x_ts_m, y_ts_m, _ = static_dataset(feature_type, 'validation', 'M', emotions,
frame_number)
#FEMALE DATASET
x_tr_f, y_tr_f, class_weight_dict_f = static_dataset(feature_type, 'train',
'F', emotions,
frame_number)
x_ts_f, y_ts_f, _ = static_dataset(feature_type, 'test', 'F', emotions,
frame_number)
#DATASET NORMALIZATION
x_tr_m = numpy.array(x_tr_m)
x_tr_m = tf.keras.utils.normalize(x_tr_m, axis=-1, order=2)
y_tr_m = numpy.array(y_tr_m)
regu = 0.0000
bias = True
drop_rate = 0.2
feature_number = 87
if feature_type == "LLF":
feature_number = 31
model = FFNN(trainable=trainable,
feature_number=feature_number,
frame_number=frame_number,
emotions=emotions,
lr=0.0001)
#FEMALE DATASET
x, y, class_weight_dict = static_dataset(feature_type, 'train', 'F', emotions,
frame_number)
x_v, y_v, _ = static_dataset(feature_type, 'test', 'F', emotions, frame_number)
#FEMALE DATASET
print(len(x))
x = x[::4]
y = y[::4]
#DATASET NORMALIZATION
x = numpy.array(x)
x = tf.keras.utils.normalize(x, axis=-1, order=2)
y = numpy.array(y)
#cass_weight_dict = weight_class('train',emotions,'M')
#callback
plot_losses = PlotLosses()
earlystop = EarlyStopping(monitor='val_acc', min_delta=0.0001, patience=10, \
verbose=1, mode='auto')
trainable = 'True'
#emotions = ['ang','dis','exc','fea','fru','hap','neu','oth','sad','sur','xxx']
emotions = ['ang','exc','neu','sad']
size_batch2 = 8
frame_number = 1
x,y,class_weight_dict = static_dataset('train','M',emotions,frame_number)
x_val,y_val,class_weight_dict_test = static_dataset('validation','M',emotions,frame_number)
x_test,y_test,class_weight_dict_test = static_dataset('test','M',emotions,frame_number)
x = tf.keras.utils.normalize(x, axis=-1, order=2)
#search parameters for hyperparameter optimization
#batch_sizes = [16, 32, 64, 128, 256]
batch_sizes = [8]
#epochs = [10, 50, 100]
epochs = [300]
#optimizers = [sgd, rmsdrop, adagrad, adadelta, adam, adamax, nadam]
#learn_rates = [0.00001, 0.0001, 0.001, 0.01, 0.1, 0.2]
#learn_rates = [0.0001,0.00005,0.00001,0.000005,0.000001]
learn_rates = [0.01,0.005,0.001,0.0005,0.0001,0.00005,0.00001,0.000001,0.0000001]