Using a DCGAN for semi-supervised learning

Data is the most important part of developing any deep learning model. However, good data is often scarce and expensive to acquire. The good news is that GANs can lend us a hand in these situations by artificially producing novel training examples, in a process known as semi-supervised learning.

In this recipe, we'll develop a special DCGAN architecture to train a classifier on a very small subset of Fashion-MNIST and still achieve a decent performance.

Let's begin, shall we?

Getting ready

We won't require anything extra to access Fashion-MNIST because it comes bundled with TensorFlow. In order to display a nice-looking progress bar, let's install tqdm:

$> pip install tqdm

Let's now move on to the next section to start the recipe's implementation.

How to do it…

Perform the following steps to complete the recipe:

1. Let's start by importing the required packages:

   import...