1
$\begingroup$

My goal is to generate artificial sequences of real-valued data (e.g. time series) with GANs. Starting simple I tried to generate realistic sine-waves using a Wasserstein GAN. But even on this simple task it fails to generate any useful samples.

This is my model:

Generator

model = Sequential()
model.add(LSTM(20, input_shape=(50, 1)))
model.add(Dense(40, activation='linear'))
model.add(Reshape((40, 1)))

Critic

model = Sequential()
model.add(Conv1D(64, kernel_size=5, input_shape=(40, 1), strides=1))
model.add(MaxPooling1D(3, strides=2))
model.add(LeakyReLU(alpha=0.2))
model.add(Conv1D(64, kernel_size=5, strides=1))
model.add(MaxPooling1D(3, strides=2))
model.add(LeakyReLU(alpha=0.2))
model.add(Flatten())
model.add(Dense(1))

Is this model capable of learning such a task or should I use a different model architecture?

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.