No longer on this platform.
Québec
Joined February 2012
- Tweets131
- Following127
- Followers1.5K
- Likes643
Pinned Tweet
For the last ~2 years I've been working on an open-source GFlowNet library, conceived around & optimized for graphs. It's been a very fun project & I've had amazing help & contributions.
Anyways, I recently wrote a colab on how to use the library, enjoy! colab.research.google.com/dr…
Emmanuel Bengio retweeted
Our bodies’ molecules aren’t static so why do many AI models treat them that way?
Without considering the physical movement of molecules, many AI tools fail to robustly capture what makes drug candidates promising.
AquaGen is a generative AI framework built to close that gap: an all-atom, explicit-solvent model that runs an order of magnitude faster than traditional MD methods.
We’re excited about the implications of this work on early-stage drug discovery.
Technical report and blog below.
Emmanuel Bengio retweeted
1/ At Valence Labs, @RecursionPharma's AI research engine, we’re focused on advancing drug discovery outcomes through cutting-edge computational methods
Today, we're excited to share our vision for building virtual cells, guided by the predict-explain-discover framework 🧵
Emmanuel Bengio retweeted
Ecstatic to show off some work my brilliant colleagues and I did at @iclr_conf this year! 🚀
We address the credit assignment challenge under long trajectories in RL or GFlowNets by constructing high order actions, or “chunks”, effectively compressing trajectory lengths!
Emmanuel Bengio retweeted
We’ve released an improved version of SynFlowNet!
SynFlowNet can now handle large action spaces (up to 220k building blocks) and we experiment with training different GFlowNet backward policies. (1/3)
arXiv: arxiv.org/abs/2405.01155
code: github.com/mirunacrt/synflow…
Emmanuel Bengio retweeted
3️⃣ “Learning to Scale Logits for Temperature-Conditional GFlowNets”
Where: Poster Session 3 - Hall C 4-9 #1411
When: Wednesday, July 24th from 11:30am - 1:00pm
Paper: arxiv.org/abs/2310.02823
@folinoid
Emmanuel Bengio retweeted
Can we perform unbiased bayesian posterior inference with a diffusion model prior? We propose Relative Trajectory Balance (RTB) which allows us to directly optimize for this posterior model. We apply this to several tasks in image, language and control!🧵arxiv.org/abs/2405.20971
Feedback is of course welcome, and this GFlowNet library, github.com/recursionpharma/g…, is far from being final! There are many more features I want to add, features that exist that aren't yet in trunk, & features I haven't thought of yet, PRs welcome ;)
Emmanuel Bengio retweeted
Very excited to present SynFlowNet next week at @gembioworkshop #ICLR2024! We tackle the synthesisability issue in molecular design by training a GFlowNet with an action space of chemical reactions. Thanks to my coauthors @charlieharris01 @juleroy13 @folinoid and Pietro Lio (1/n)
Emmanuel Bengio retweeted
M2D2 is back and @folinoid will be joining @TOSSOUPrudencio as a co-host! 🤯
First talk of 2024 will feature @Rza_ozcelik and his paper “Chemical Language Modeling with Structured State Spaces”.
📅 Tune in tomorrow, Jan 30th at 11am ET.
Details here: portal.valencelabs.com/event…
Recently sat down with Parker Levesque for a fun chat, we talked about GFlowNets (of course), compilers, grad school, research, and more! Listen here: open.spotify.com/episode/7rX…
Emmanuel Bengio retweeted
Last November, we organized a workshop at @Mila_Quebec on GFlowNets. All the videos are now available online. If you want to learn more about GFlowNets and their applications, check out this blog post: mila.quebec/en/gflownets_ai/
Emmanuel Bengio retweeted
Valence Labs will be co-hosting a TechBio social with @recursionpharma and @nvidia at #NeurIPS.
Join us on Wed, Dec 13th. RSVP here: lu.ma/5wx660v2
Our team will also be at #NeurIPS throughout the week. See below for a summary of our papers👇
Emmanuel Bengio retweeted
@dhanya_sridhar, @chandlersquires and I are running a talk series on Causality, Abstraction, Reasoning and Extrapolation (CARE).First talk is at 11EST tomorrow by @quanhanxi on "The Statistical Structure of Identifiable Generative Models".All are welcome! portal.valencelabs.com/event…
Emmanuel Bengio retweeted
Thrilled to announce the first annual Reinforcement Learning Conference @RL_Conference, which will be held at UMass Amherst August 9-12! RLC is the first strongly peer-reviewed RL venue with proceedings, and our call for papers is now available: rl-conference.cc.
Emmanuel Bengio retweeted
From November 8 to 10, Mila will be holding its first workshop on generative flow networks (GFlowNets).
Save this link and join us online tomorrow as of 9:30am : youtube.com/@mila-quebecarti…
Emmanuel Bengio retweeted
Valence Labs is actively hiring across multiple roles!
We’re looking for Research Engineers, Research Scientists, a DevOps Engineer, and a Community Manager.
Join us in pushing the frontier of AI research in drug discovery.
valencelabs.com/careers
Emmanuel Bengio retweeted
There are also many opportunities for Master's and PhD students!
Apply for the Machine Learning Research Intern role or submit a research proposal to be considered for the Valence scholarship.
You can find more details on the scholarship here: valencelabs.com/careers?gh_j…
Emmanuel Bengio retweeted
(1/8) We propose a rigorous framework for investigating Molecular Out-of-Distribution (MOOD) performance in real-world applications of molecular modeling.
It’s time to close the validation-deployment gap! 🎯
Paper: chemrxiv.org/engage/chemrxiv…
Experiments: github.com/valence-labs/mood…