Translating music to predict a musician’s body movements

Long story short: animating personalized avatar with the help of neural networks.

https://research.fb.com/translating-music-to-predict-a-musicians-body-movements/

#lstm #cvpr2018 #facebook

This is a day to remembered. #OpenAI 's team of five neural networks, OpenAI Five, has started to defeat amateur human teams (including a semi-pro team) at Dota 2:
https://blog.openai.com/openai-five/

It is important, because Dota2 is a way more complicated game than Chess or Go, where #AI has already surpassed human players.

#rl #reinforcementlearning #dl #dota2 #lstm

​​Automatically Generating Comments for Arbitrary Source Code

Automatically generating code comments directly from source code using an LSTM. Works with multiple languages. Can’t wait to JetBrains discovering it.

Link: https://www.twosixlabs.com/automatically-generating-comments-for-arbitrary-source-code/

#NLP #CS #coding #LSTM

​​How Uber predicts prices

Engineering Uncertainty Estimation in Neural Networks for Time Series Prediction at Uber

Link: https://eng.uber.com/neural-networks-uncertainty-estimation/

#RNN #LSTM #Uber

Engineering Uncertainty Estimation in Neural Networks for Time Series Prediction at Uber

More complex article on #TS forecasting from #Uber team.

Link: https://eng.uber.com/neural-networks-uncertainty-estimation/

#RNN #LSTM #Uber

​​LSTM on Amazon Food Reviews using Google Collaboratory

Article describing how to build easy and small #LSTM network to predict review score based on its text, using #GoogleCollab. This is an #entrylevel post, useful if you have medium experience in #NLP.

Link: https://medium.com/@theodoxbolt/lstm-on-amazon-food-reviews-using-google-collaboratory-34b1c2eceb80

#novice

​​Single Headed Attention RNN

tl;dr: stop thinking with your (#attention) head :kekeke:

* obtain strong results on a byte level #languageModeling dataset (enwik8) in under 24 hours on a single GPU (12GB Titan V)
* support long-range dependencies (up to 5000 tokens) without increasing compute time or memory usage substantially by using a simpler attention mechanism
* avoid the fragile training process required by standard #Transformer models such as a long warmup
* back off toward a standard #LSTM allowing you to drop retained memory states (needed for a Transformer model) if memory becomes a major constraint
* provide a smaller model that features only standard components such as the LSTM, single-headed attention, and feed-forward modules such that they can easily be productionized using existing optimized tools and exported to various formats (i.e. ONNX)

paper: https://arxiv.org/abs/1911.11423
code: https://github.com/Smerity/sha-rnn
tweet: https://twitter.com/Smerity/status/1199529360954257408?s=20