Word
mode collapse
Definition
n. a problem in generative AI where the model gets stuck producing only one or two repetitive answers instead of varied outputs.
n. a failure mode in generative models (notably GANs) where the generator learns to map distinct latent inputs to a severely restricted subset of data distribution modes.
Examples
“The face generator suffered from mode collapse and output only pictures of the same smiling woman.”
“Wasserstein loss functions were introduced specifically to prevent mode collapse during adversarial training.”
“Diffusion models largely superseded generative adversarial networks because their progressive denoising objectives are fundamentally immune to mode collapse.”
Examples
simple
“The face generator suffered from mode collapse and output only pictures of the same smiling woman.”
contextual
“Wasserstein loss functions were introduced specifically to prevent mode collapse during adversarial training.”
complex
“Diffusion models largely superseded generative adversarial networks because their progressive denoising objectives are fundamentally immune to mode collapse.”
Real-World Examples
“Mode collapse, characterised by the generator producing limited variations of samples that fail to cover the full spectrum of the target distribution, often arises due to gradient instability in the generator.” “Mode collapse occurs in GANs when the generator model fails to produce a diverse range of outputs that accurately capture every aspect of diversity present in the real data distribution, which reduces the efficacy of the synthetic data.” Etymology
From statistical mode (Latin modus, 'measure, manner', the most frequent value) + collapse (Latin collabi, 'to fall together'); introduced by Ian Goodfellow et al. in 2014.
Etymology adapted from Wiktionary, available under CC BY-SA 4.0.
Domains
Scan code
englishreference.com/q/mode-collapse