Head of AI Research in a fintech company. Decision Making in the Wild. Opinions are my own.
Joined May 2011
- Tweets589
- Following239
- Followers1.2K
- Likes1.5K
Pinned Tweet
This year we managed to do the impossible. We launched the AI research department, and we launched it noticeably – with recognition from the world's leading AI conferences: ICML and NeurIPS (both spotlights). But I want to believe that this is just the beginning...
Sergey Kolesnikov retweeted
1/7 In-context learning (ICL) is poised to revolutionise NLP, but its success hinges on our ability to process long sequences. Recently, @simran_s_arora et al. showcased advancements in Linear Transformers and proposed Based. But what if we could push its boundaries further?
Sergey Kolesnikov retweeted
Our first stable release and full paper preprint for XLand-MiniGrid is done, check it out! Compared to the workshop version, we have significantly redesigned the library, multi-GPU baselines and standardized benchmarks with millions of unique tasks.
github.com/corl-team/xland-m…
Sergey Kolesnikov retweeted
In-Context RL for Variable Action Spaces
ICRL is a promising direction to build Foundational Decision-Making Models. But adaptation to new action spaces is a problem.
We propose Headless Algorithm Distillation (@MishaLaskin) to address it.
arxiv.org/abs/2312.13327
Sergey Kolesnikov retweeted
Which data-collection strategies enable In-Context Reinforcement Learning?
You need either RL training trajectories or supervision with optimal actions. But what if we had a demonstrator policy, could we use it to enable ICRL? We show the answer is yes
arxiv.org/abs/2312.12275
Sergey Kolesnikov retweeted
🔥 Imagine if you could train Meta-RL agents for 1 TRILLION transitions under 40 hours?
We present XLand-MiniGrid — JAX-accelerated meta-reinforcement learning environments inspired by XLand (@FeryalMP) and MiniGrid (@Love2Code).
code: github.com/corl-team/xland-m…
Sergey Kolesnikov retweeted
NetHack is arguably one of the most challenging games for humans and even more for RL algorithms. Maybe, offline RL could help? Time will reveal.
To bootstrap the practitioners, we release Katakomba — Tools and Benchmarks for Data-Driven NetHack.
github.com/tinkoff-ai/katako…
Sergey Kolesnikov retweeted
Interested in offline and offline-to-online RL 🫶? Check out new major release of Clean Offline Reinforcement Learning library:
🤖 Offline: 10 algorithms, 30 datasets benchmarked
🦾 Offline-to-Online: 5 algorithms, 10 datasets benchmarked
github.com/tinkoff-ai/CORL
📢 Exciting research from our team! We explored the power of seemingly minor design choices in offline RL by applying them to an established minimalistic baseline developed by @shaneguML. The outcome? Just follow this 🧵
There were a lot of algorithmic innovations in offline RL recently, along with a silent evolution of minor design choices.
What if we applied these seemingly minor modifications to an established minimalistic baseline by @shaneguML? Turns out, gains are enormous.
Exciting news: Our paper has been accepted at ICML! Our work focuses on improving the reliability of offline RL algorithms and tackling overfitting through an anti-exploration bonus. And the best part? SAC-RND challenges SOTA results with a single network, no ensembles required!
For a full list of our publications, check my unofficial records 😅
notion.so/scitator/TRS-Paper…
Optimizing accuracy is not a problem if you are EXACT 🤘
In our new paper 🚀 we optimize accuracy via gradient descent! The work, called "EXACT: How to Train Your Accuracy", will be presented at the TAG-ML workshop during #ICML2022 🙃
Paper: arxiv.org/pdf/2205.09615.pdf
Poster: drive.google.com/file/d/1ZBO…
Enjoy!)
You don't need a TPU cluster to count a budget 🤯
Join our EOP talk (@vladkurenkov) in room 307 in 2 hours!
@icmlconf spotlight, #ICML2022
Are you ready for the upcoming ICML spotlight? 🤯
kudos to @vladkurenkov
Extremely pleased to announce that our paper “Showing Your Offline Reinforcement Learning Work: Online Evaluation Budget Matters” was accepted to ICML 2022 (Spotlight)!
tinkoff-ai.github.io/eop
(1/N)
If you're at ICLR now, check Generalizable Policy Learning in the Physical World workshop tomorrow. We will present Prompts and Pre-Trained Language Models for Offline Reinforcement Learning and will be happy to share its current improvements. We have some 😉
Sergey Kolesnikov retweeted
Check out our blog post for a thorough explanation of how brainchop.org was made and the principles behind its work. With @MMasoud2021 @FarfallaHu @Entodi @Kevin_C_Wang @Scitator
#neuroimaging #brainresearch #medicalresearch #MRI #MadeWithTFJS
trendscenter.org/in-browser-…
I was experimenting with Self-Supervised Learning recently, so if you are interested in easy to go implementations for Barlow-Twins, BYOL, SimClR, or Supervised contrastive - check out the repo: bit.ly/3H9OgUa
Comparison results included!
Long time no see, my friends!
I am thrilled to present you with a new year update of the Catalyst - PyTorch high-level API to accelerate your R&D.
Bunch of improvements and simplifications, check out updated examples for more:
bit.ly/3v4nw53
Interested in Offline RL?
We have recently updated our work on Online Evaluation Performance: bit.ly/3rNljZP
Open for your questions in the thread!
