Pandaemonium Architecture 6.0

In this course

Odena is a foundational researcher in GAN stability and failure modes, whose 2016 work documented characteristic artifacts in generative adversarial networks—particularly "local artifacts" and texture-blob pathologies that became reference points for subsequent work on training dynamics. The readings invoke Odena's catalogue of failure modes as a baseline against which to measure new instabilities emerging at large scale, using this typology to argue that contemporary scaling introduces distinct problems requiring new analytical frameworks beyond toy-problem diagnostics.

Mentioned in 1 reading

Andrew Brock

et al., LARGE SCALE GAN TRAINING FOR HIGH FIDELITY NATURAL IMAGE SYNTHESIS

4

Of particular relevance to our work is Spectral Normalization (Miyato et al., 2018), which enforces Lipschitz continuity on D by normalizing its parameters with running estimates of their first singular values, inducing backwards dynamics t…Much previous work has investigated GAN stability from a variety of analytical angles and on toy problems, but the instabilities we observe occur for settings which are stable at small scale, necessitating direct analysis at large scale. We…We note that some failure modes of our partially-trained models are distinct from those previously observed. Most previous failures involve local artifacts (Odena et al., 2016), images consisting of texture blobs instead of objects (Saliman…

+1 more

Appears alongside

People mentioned in the same passages — sorted by co-occurrence weight.