GANs are comprised of two neural networks: a generator and a discriminator. They are able to generate new, synthetic data. The generator outputs new instances of the data, while the discriminator determines whether each instance of the data that is fed to it belongs to the training dataset.
The following screenshot gives an illustration of the output from a GAN on the MNIST and Toronto Face datasets. In both cases, the images on the far-right side of the grid are the true values and the others are generated by the model:
The source for this image can be found at: https://arxiv.org/pdf/1406.2661.pdf
Let's consider this further in the context of using the MNIST dataset, where the goal of the GAN is to generate similar images of handwritten digits. The role of the generator in the network is to create new synthetic images. These images are then passed to...
