Pandaemonium Architecture 6.0

In this course

Miyato is known for developing spectral normalization, a crucial regularization technique that stabilizes GAN discriminators by constraining Lipschitz continuity through singular value normalization. His work appears throughout the readings as a foundational method for training conditional GANs, particularly in how class information is injected into discriminator networks through normalized embeddings and feature conditioning. The course draws on Miyato's contributions to show how mathematical constraints on network parameters—singular values, cosine similarity conditioning—became essential tools for making generative models controllable and reliable.

Mentioned in 1 reading

Andrew Brock

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

8

A Generative Adversarial Network (GAN) involves Generator (G) and Discriminator (D) networks whose purpose, respectively, is to map random noise to samples and discriminate real and generated samples. Formally, the GAN objective, in its ori…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…Dumoulin et al. (2017) modify the way class conditioning is passed to G by supplying it with class-conditional gains and biases in BatchNorm (Ioffe & Szegedy, 2015) layers. In Miyato & Koyama (2018), D is conditioned by using the cosine sim…

+5 more

Appears alongside

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