🍀 Challenge: Define the Discriminator

The task will test your knowledge on building the discriminator class for the conditional GAN.

We'll cover the following

We have already discussed the Fashion MNIST dataset here.

Problem statement

A partial Discriminator class for the Fashion MNIST dataset is provided to you. Your task is to define the:

  • Network architecture

    • The input layer should have 784+10 nodes.

    • The hidden layer 1 should have 200 nodes with the LeakyReLU activation applied to the output. The hidden layer output should be normalized.

    • The output layer should have 1 node with the Sigmoid activation function applied to the output.

  • Loss function as BCE.

  • Optimiser as Adam. The learning rate should be 0.0001.

Get hands-on with 1200+ tech skills courses.