Generative Adversarial Nets

Generative Adversarial Nets asks how a model can learn to create convincing new samples without an explicit likelihood or a Markov chain. Read it to see how a generator and discriminator turn that problem into a two-player game, why the ideal judge eventually answers one half, and why the elegant equilibrium does not make practical training automatically stable.

Reading focus: How a generator maps noise into samples while a discriminator learns to separate generated examples from real data and sends a useful gradient back. Why the minimax objective reaches its ideal distribution-space optimum when generated and real distributions match and the best discriminator returns one half. Why alternating updates, the non-saturating generator loss, synchronization, implicit density, discrete data, collapse, and limited evaluation keep the theorem separate from practical training.

NeurIPS 2014. Goodfellow, Pouget-Abadie, Mirza, Xu, Warde-Farley, Ozair, Courville, and Bengio. 35 min read, easy difficulty.