def __init__(self, filters=1, kernel_size=80, rank=1, strides=1, padding='valid',
data_format='channels_last', dilation_rate=1, activation=None, use_bias=True,
fsHz=1000.,
fc_initializer=initializers.RandomUniform(minval=10, maxval=400),
n_order_initializer=initializers.constant(4.),
amp_initializer=initializers.constant(10 ** 5),
beta_initializer=initializers.RandomNormal(mean=30, stddev=6),
bias_initializer='zeros',
**kwargs):
super(Conv1D_gammatone, self).__init__(**kwargs)
self.rank = rank
self.filters = filters
self.kernel_size_ = kernel_size
self.kernel_size = conv_utils.normalize_tuple(kernel_size, rank, 'kernel_size')
self.strides = conv_utils.normalize_tuple(strides, rank, 'strides')
self.padding = conv_utils.normalize_padding(padding)
self.data_format = conv_utils.normalize_data_format(data_format)
self.dilation_rate = conv_utils.normalize_tuple(dilation_rate, rank, 'dilation_rate')
self.activation = activations.get(activation)
self.use_bias = use_bias
self.bias_initializer = initializers.get(bias_initializer)
self.fc_initializer = initializers.get(fc_initializer)
self.n_order_initializer = initializers.get(n_order_initializer)
self.amp_initializer = initializers.get(amp_initializer)
self.beta_initializer = initializers.get(beta_initializer)
self.input_spec = InputSpec(ndim=self.rank + 2)
self.fsHz = fsHz
self.t = tf.range(start=0, limit=kernel_size / float(fsHz),
delta=1 / float(fsHz), dtype=K.floatx())
self.t = tf.expand_dims(input=self.t, axis=-1)
def __init__(self, vocab_size, embed_dim, hidden_size=256, log_dir="./logs"):
self.log_dir = log_dir
init_embed = initializers.RandomUniform(minval=-0.1, maxval=0.1, seed=None)
init_lstm_norm = 1 / np.sqrt(hidden_size)
init_lstm = initializers.RandomUniform(minval=-init_lstm_norm, maxval=init_lstm_norm)
self.model = Sequential()
self.model.add(
Embedding(vocab_size, embed_dim, input_length=None, name="embedding", embeddings_initializer=init_embed)
)
self.model.add(Dropout(0.2))
self.model.add(
Bidirectional(
CuDNNLSTM(
hidden_size, return_sequences=True, kernel_initializer=init_lstm, recurrent_initializer=init_lstm
)
)
)
self.model.add(Dropout(0.3))
self.model.add(
Bidirectional(
CuDNNLSTM(
hidden_size, return_sequences=True, kernel_initializer=init_lstm, recurrent_initializer=init_lstm
)
)
)
self.model.add(Dropout(0.3))
self.model.add(Bidirectional(CuDNNLSTM(embed_dim // 2, return_sequences=True)))
self.model.add(Dropout(0.3))
self.model.add(
TimeDistributed(
TiedEmbeddingsTransposed(tied_to=self.model.get_layer(name="embedding"), activation="softmax")
)
)
self.optim = optimizers.SGD(lr=1.0, decay=1e-4, momentum=0.9, nesterov=True, clipnorm=0.5)
self.model.compile(loss="sparse_categorical_crossentropy", optimizer=self.optim, metrics=["accuracy"])
print(self.model.summary())
tot_mae = []
tot_acc = []
for i in ["1", "2", "3", "4", "5"]:
model = Sequential()
X_train, y_train, X_test, y_test = read_data_100k_gen_dataset(i)
model_scalar = StandardScaler()
model_scalar.fit(X_train)
X_train = model_scalar.transform(X_train)
X_test = model_scalar.transform(X_test)
initVar = initializers.RandomUniform()
inputDimen = np.shape(X_train)[1]
outputDimen = 5
model.add(
Dense(units=1000,
input_dim=inputDimen,
trainable=False,
kernel_initializer=initVar))
model.add(Activation('linear'))
model.add(Dense(units=2000))
model.add(Activation('relu'))
model.add(Dense(units=outputDimen))