Advanced · Generative AI

Mode Collapse

Visual diagram · (in preparation) · Math · (in preparation) · Worked example · 3 difficulty levels.

TL;DR. A GAN failure where the generator produces only limited variety, ignoring the full diversity of training data.

Technical Definition

A GAN failure where the generator produces only limited variety, ignoring the full diversity of training data.

How it works

Mode collapse occurs when the generator finds a few outputs that fool the discriminator and keeps producing only those. Solutions include Wasserstein GAN, minibatch discrimination, progressive training, and spectral normalization.

Related Concepts

• Deep Learning — A subset of machine learning using neural networks with many layers to learn hierarchical representations from large datasets.
• Loss Function — A mathematical function that measures how far the model's predictions are from the actual values, guiding the learning process.
• Generative Adversarial Network (GAN) — Two neural networks — a generator and discriminator — compete against each other to produce increasingly realistic synthetic data.