GoogleAI has just released a new library for training machine learning models with (differential) privacy for training data! Very excited to share it with the research community and we look forward to your contributions! Check it out
This is how little code it takes to implement a siamese net using @fastdotai and @pytorch. I share this because I continue to be amazed.
We have created and are now open-sourcing Nevergrad, a Python3 library that makes it easier to perform gradient-free optimizations used in many machine learning tasks
We train a classifier to predict whether a paper should be rejected based solely on the visual appearance of the paper. Results show that our classifier can safely reject 50% of the bad papers while wrongly reject only 0.4% of good papers. The author of this paper-classifier ran their model on their paper describing this paper-classifier and it suggested a strong reject. 🤣
Really cool design of a spiking NN that appears to actually outperform LSTM and GRU!
Deep Networks Incorporating Spiking Neural Dynamics https://arxiv.org/abs/1812.07040
JAX is NumPy on the CPU and GPU with autodiff and JIT compilation. It’s just been released, try it out!
As demonstration, we fit a 10-billion parameter ‘Bayesian Transformer’ on 512 TPUv2 cores, which replaces attention layers with their Bayesian counterpart. 🔥
Stunning face generation results from NVidia Research! They do not exist on planet Earth. https://arxiv.org/pdf/1812.04948.pdf
