Implementing Essential Deep Learning Modules - Week 4

We have finally wrapped up with Phase I and along with it, a number of key goals.

The GAN code is finally merged, and the DCGAN code has been debugged for the MNIST dataset. Now the only thing remaining to be done is to debug the test for the CelebA dataset, and the PR can be merged thereafter. Also, then we can upload the code to the models repository, though for now all the code is hosted within mlpack itself, as this is a secondary goal.

This week we got some promising results with the GAN code on the 10,000 image subset and the full 70,000 image set of MNIST for you to feast your eyes on:

10,000 Images

70,000 Images

For the next week, we will be focussing on implementing the support for batch sizes and optimizer separation for GANs.

Vale!