03 / Generative models · Unsupervised learning
Generative adversarial networks
A generator and a discriminator, learning from each other through adversarial training.
ADVERSARIAL TRAINING03
Noise → generate → discriminateThe question
How can a model learn to generate samples by competing with another model?
The approach
I built a basic generative adversarial network. The generator takes random noise as its input. The discriminator learns to distinguish real samples from generated ones, providing the feedback used to train the generator.
How it works
Start with noise
The generator maps random input to a generated sample. It learns through feedback from the discriminator.
Train both sides
The discriminator learns from real and generated samples. The two models train against each other, making the training process the focus of this experiment.
Next project / 04