CS5720 - Week 8
Slide 146 of 160

GAN Architecture: Generator vs Discriminator

🎨

The Generator

  • 📥 Input: Random noise vector (z)
  • 🏗️ Architecture: Deconvolutional layers
  • 📤 Output: Synthetic data samples
  • 🎯 Goal: Fool the discriminator
  • 📚 Training: Learn data distribution
🔍

The Discriminator

  • 📥 Input: Real or fake samples
  • 🏗️ Architecture: Convolutional layers
  • 📤 Output: Probability [0, 1]
  • 🎯 Goal: Distinguish real from fake
  • 📚 Training: Binary classification

The Adversarial Battle

🎲
Random Noise
z ~ N(0, 1)
→
🎨
Generator
G(z)
→
🔍
Discriminator
D(G(z))
The Minimax Game: minG maxD V(D,G) = E[log D(x)] + E[log(1 - D(G(z)))]
← Previous Next →
Prepared by Dr. Gorkem Kar