Loading your learn pack…
Read the summary, then start training!
GANs train a generator and discriminator in a zero-sum minimax game, using the discriminator’s feedback as an indirect learning signal.
A generative adversarial network (GAN) is a framework for generative AI in which two neural networks compete in a zero-sum game: a generator and a discriminator. The generator produces synthetic samples from a latent space, while the discriminator tries to distinguish generated (fake) samples from real data. Training proceeds indirectly: the generator is updated to “fool” the discriminator rather than to directly minimize a distance to any specific target sample. Mathematically, the original GAN can be described as a minimax game. For a reference data distribution \(\mu_{ref}\) and generator distribution \(\mu_G\), the discriminator \(D\) outputs a probability that an input is real. The objective \(L(\mu_G,\mu_D)\) is maximized by the discriminator and minimized by the generator. Intuitively, the discriminator provides a learning signal about how realistic the generator’s outputs are, and as the generator improves, the discriminator’s task becomes harder. In practice, GANs are trained by alternating updates: first training the discriminator on real samples and current generated samples, then updating the generator based on how well it succeeds at producing samples the discriminator misclassifies as real. This adversarial setup yields an implicit generative model (it does not explicitly model likelihoods), and under the original formulation the optimal discriminator relates the game to a divergence between distributions (Jensen–Shannon divergence), giving a theoretical intuition for why matching the data distribution is the goal.
GANs train a generator and discriminator in a zero-sum minimax game, using the discriminator’s feedback as an indirect learning signal.
The generator aims to match the reference data distribution, while the discriminator aims to output high scores for real samples and low scores for generated ones.
GANs are implicit generative models: they generate samples in one forward pass and typically do not require explicit likelihood computation.
A neural network that maps latent noise to synthetic samples, inducing a model distribution \(\mu_G\).
A neural network that outputs the probability an input is real versus generated, used to form the adversarial learning signal.
The training objective where the discriminator maximizes and the generator minimizes the same loss function.
A generative model that does not explicitly define or compute a likelihood function for data, but instead learns to sample from a learned distribution.
A divergence between probability distributions that appears in the theory of the original GAN via the optimal discriminator.
“Can you explain what "GANs train a generator and discriminator in a zero-sum minimax game, using the discriminator’s feedback as an indirect learning signal." means in simple terms?”