AI Workshop: Hands-on with GANs using Dense Neural Networks

Course details

If you’re looking for hands-on AI practice, this workshop-style coding course was designed for you. Join instructor Janani Ravi as she shows you how to build and train generative adversarial networks (GANs). Explore the core components of GANs, including how to set up the virtual environment, run the notebook server, instantiate the PyTorch Dataset, DataLoader, and more. Janani covers the basics of standalone training of adversaries, training GANs, and visualizing results. Janani also discusses how to address common problems associated with GANs and mitigate them effectively throughout the training process.

Concepts

Introducing Generative Modeling

Understanding generative modeling

Introducing Generative Adversarial Networks (GANs)

Course outline and prerequisites
Set up the virtual environment and run the notebook server
Introducing GANs
Instantiating the dataset and data loader
Viewing training data

Stand-Alone Training of Adversaries

Big picture overview of a GAN
Training the adversaries
The generator architecture
The discriminator architecture
Understanding the generator and discriminator outputs
Stand-alone training of a discriminator as a classification model
Stand-alone training of a generator

Training GANs

Computing losses for generators and discriminators
Understanding the minimax loss function
Setting up GAN training
Visualizing GAN training results
Problems with GANs and potential mitigations

Conclusion

Summary and next steps

AI Workshop: Hands-on with GANs using Dense Neural Networks

Authors

Janani Ravi