Yazhe Li
@yazhe_li
Followers
288
Following
14
Statuses
27
Research engineer @DeepMind; PhD student at Gatsby Computational Neuroscience Unit
London, England
Joined December 2017
It really isn't a lot of money to ask for, but could transform someone's Christmas. A hot meal, a kind word – those simple things could bring comfort and joy to someone spending the holidays alone.
0
0
0
Huge thanks to my advisors, collaborators, examiners, family, friends and everyone who supported me along this incredible journey!
Congratulations Dr @yazhe_li on a successful thesis defence! Thank you to the excellent thesis examiners @y_m_asano and Marc Deisenroth!
1
0
13
RT @ArthurGretton: Testing for conditional independence is notoriously difficult. Here's o̶n̶e̶ three weird tricks to: 👾 make testing…
0
17
0
RT @ArthurGretton: Turn your HSIC dependence statistic into a conditional dependence statistic with CIRCE! Learn NN features that are indep…
0
14
0
Wondering if model-based RL helps generalization? Our new work provides some insights.
Model-based RL promises generalization by design, but do MBRL agents like MuZero generalize better than model-free, and benefit from self-supervision? The answer is yes! MuZero+SSL gets SotA on Procgen with 10x less data, implicit meta-RL on MetaWorld:
0
1
10
RT @ArthurGretton: Your contrastive self-supervised learning is secretly a kernel method.... with @yazhe_li @rmnpog…
0
36
0
RT @OriolVinyalsML: MuZero removed simulators in MBRL vs AlphaGo. VQ Models for Planning generalize to partial observable & stochastic envi…
0
45
0
I have to say this accurately describes my feeling regarding the recent events.
@dlowd @AnimaAnandkumar Every fight has two sides. To be honest, I found both parties verbally abusive, such that so far I hesitate to say anything. Yes, it is unpleasant environment so to I do not feel to express my view. so the freedom of speech is diminished here.
0
0
3
Very impressive result and a huge improvement to CASP13 model. Congratulations to the protein folding team!
In a major scientific breakthrough, the latest version of #AlphaFold has been recognised as a solution to one of biology's grand challenges - the “protein folding problem”. It was validated today at #CASP14, the biennial Critical Assessment of protein Structure Prediction (1/3)
0
0
2
Nice! 👍 This note provides much deeper insight and is much better presented than my presentation at Gatsby.
New post (after almost a year): Notes on Causally Correct Partial Models by @DaniloJRezende and colleagues
0
0
1
After saying no one (including medical professionals) needs mask, only old people with health condition would die, this one is truly eye-opening. The Guardian: Michael Gove appears to blame China over lack of UK coronavirus testing. via @GoogleNews
0
0
0
@random_walker Never realised this, but now think about it, it is a good lesson to learn. Cannot remember how many times my name was spelled incorrectly, don't even mention the pronunciation. Probably need a solution that is not changing name.
0
0
0
RT @GatsbyUCL: Interested in doing a PhD with us? Applications for our 2020 PhD programme in Theoretical and Computational Neuroscience and…
0
93
0