Dynamically Typed

Not-so-BigGAN

DT #48 — September 13, 2020

When I covered BigGAN in February 2019 (DT #6), its image generation results were very impressive — but the model was also incredibly expensive to train, requiring a cluster of hundreds of TPUs. Now, just a year and a half later, Han et al. (2020) introduced not-so-BigGAN: close-enough image quality trained on just 4 Tesla-V100 GPUs — an order of magnitude less compute. This speed of this progress is amazing.

ML Research

This section of Dynamically Typed covers recent models, datasets, and tools for machine learning research.

Join 325+ others and subscribe to get DT in your inbox every second Sunday — 76 issues and counting!

Or check out recent DT issues first:

DT #76: Dynamically Typed Hiatus
DT #75: OpenAI's book summaries for the alignment problem, Translatotron 2, and AI-generated movie posters
DT #74: Apple's privacy-focused facial recognition, DeepMind's multimodal Perceiver IO, and sea ice forecasting with IceNet

ML Research

Recent models, datasets, and tools for machine learning research.

14 stories 86 links