Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Lukasz Kaiser at AI Frontiers: One Model to Learn It All


Published on

Deep learning yields great results across many fields, from speech recognition, image classification, to translation. But for each problem, getting a deep model to work well involves research into the architecture and a long period of tuning. We present a single model that yields good results spanning multiple domains. This single model is trained concurrently on ImageNet, multiple translation tasks, image captioning, a speech recognition corpus, and an English parsing task. We achieved state-of-the-art performance while training much quicker and generating long coherent pieces, even on the scale of full Wikipedia articles. Our new architectures improve the ability to generate both text and images

Published in: Data & Analytics
  • Be the first to comment

Lukasz Kaiser at AI Frontiers: One Model to Learn It All

  1. 1. ● ● ● ● ● ●
  2. 2. ● ● ●
  3. 3. Model Type % unrecognized (max = 50%) ResNet 4.0% Superresolution GAN (Garcia’16) 8.5% PixelRecursive (Dahl et al., 2017) 11% Image Transformer 36.9%
  4. 4. ● ● ● ● ● ●