Publicações

Today we announced our partnership with OpenMined to offer developers a new series of #PyTorch privacy and machine learning courses:
- Privacy and Society
- Foundations of Private Comp.
- Enterprise Federated Learning
- Federated Learning on Mobile...
https://ai.facebook.com/…/facebook-ai-openmined-partner-on-…

Ver mais
A imagem pode conter: ?texto que diz "?OpenMined ? PyTorch FACEBOOK Al?"?

What does an AI model “understand” and why? A long-held belief is there are easy-to-interpret neurons -- or “class selective” neurons. For instance, finding neurons that only activate cat images are the “cat” part of the network.

Surprisingly, we found new evidence that demonstrates class selective neurons can actually impair the performance of deep neural networks. Learn more: http://ow.ly/jVhU50C5aR3

A imagem pode conter: texto que diz "Research in Brief baseline 50 54 40 ??e?n??e 30 sa 20 10 lie enesa 50 52 o 0.0 46 0.2 0.4 0.6 0.2 0.4 0.6 0.8 Mean class selectivity network 0.8 1.0"
Ver tudo
Vídeos
We’re releasing fairmotion, a library to help AI researchers use motion capture data for computer graphics and robotics. It provides tools to load, process and visualize motion, and demonstrates its utility with example tasks. We’ve used the library in our work presented at SIGGRAPH 2020 on controlling diverse behaviors for physically simulated characters. Get the code and paper: http://github.com/facebookresearch/fairmotion https://research.fb.com/publications/a-scalable-approach-to-control-diverse-behaviors-for-physically-simulated-characters/
350
16
Introducing Opacus: A high-speed library for training PyTorch models with differential privacy
171
5
Retrieval Augmented Generation: Streamlining the creation of intelligent natural language processing models
225
7
Ver tudo