Build Basic Generative Adversarial Networks (GANs) Course Syllabus

Overview: This course provides a comprehensive introduction to Generative Adversarial Networks (GANs), blending theoretical foundations with hands-on implementation using PyTorch. Over approximately 28 hours, learners will progress through foundational and advanced GAN architectures, gaining practical experience in building, training, and controlling generative models. The curriculum is designed for professionals seeking to deepen their understanding of generative AI with flexible scheduling to accommodate working learners.

Module 1: Intro to GANs

Estimated time: 5 hours

• Understand real-world applications of GANs
• Explore the fundamental components: generator and discriminator
• Learn the adversarial training process and min-max loss
• Build your first GAN using PyTorch

Module 2: Deep Convolutional GANs (DCGANs)

Estimated time: 6 hours

• Study the architecture of DCGANs
• Implement convolutional layers for image generation
• Apply batch normalization and transposed convolutions
• Train DCGANs on image datasets and evaluate performance

Module 3: Wasserstein GANs with Gradient Penalty (WGAN-GP)

Estimated time: 8 hours

• Identify common GAN training challenges such as mode collapse
• Understand the theoretical basis of Wasserstein distance
• Implement WGANs with gradient penalty for stable training
• Compare WGAN-GP performance with standard GANs

Module 4: Conditional GANs & Controllable Generation

Estimated time: 9 hours

• Learn the concept of conditioning in GANs
• Develop Conditional GANs to generate data from specific categories
• Integrate class labels into generator and discriminator networks
• Apply controllable generation for targeted outputs

Prerequisites

• Programming experience in Python
• Familiarity with deep learning frameworks such as PyTorch
• Basic understanding of machine learning concepts

What You'll Be Able to Do After

• Understand the core components and training dynamics of GANs
• Implement and train various GAN architectures including DCGANs and WGAN-GPs
• Build conditional GANs for category-specific data generation
• Use PyTorch to develop and experiment with generative models
• Apply GANs in computer vision and synthetic data generation tasks