
Machine learning PhD @MIT | Researcher & trader @twosigma @wehrtyou @jouhouken
Joined December 2019
- Tweets36
- Following315
- Followers952
- Likes283
Keyulu Xu retweeted
We’re very excited to share BoltzGen — a biomolecular binder design model that matches state-of-the-art folding and turns that capability into generalizable binder design!
Unlike prior tools, BoltzGen spans nanobodies, miniproteins, peptides, and cyclic scaffolds and designs binders for proteins, nucleic acids, small molecules, and more. A declarative specification language (binding sites, covalents, structure groups, secondary structure, design masks) lets you steer generation toward precise objectives. With multiple wet-lab partners, we validated across diverse targets. On 9 novel protein targets (no similar bound complexes in PDB), testing ≤15 designs/target yielded nanomolar binders on 6/9 for nanobodies (67%), mirrored by protein designs. We also show functional peptide designs (including disordered regions) and support for challenging modalities like disulfide-stabilized peptides and small-molecule binders.
The entire stack is open-source (MIT): weights, training & inference, and a full design pipeline.
🚀 Model & code: lnkd.in/e5zKCDsy
🤗 Slack community: lnkd.in/ecmfsNJ3
🧠 Blog: boltz.bio/boltzgen
📄 Manuscript: lnkd.in/eXteVX4c
Join us for live presentations, demos, and discussions: • MIT (Cambridge) — Thu Oct 30: luma.com/7474iho2• London — Thu Nov 6: luma.com/l2zgvfwt It’s been a huge pleasure working with the Boltz team, and special thanks to @HannesStaerk for leading the project!
Original post from Hannes:
nitter.cf/HannesStaerk/status/19…
Our paper shows neural networks can learn surprisingly good representations with noisy labels. Pre-training and fine tuning with a good architecture structure is the key #Neurips2021
arxiv.org/abs/2012.12896
#Neurips2021 How does a Neural Network's Architecture Impact its Robustness to Noisy Labels?
arxiv.org/abs/2012.12896 (Arxiv version will be updated soon)
Joint work with @mozhi_zhang @KeyuluXu @johnpdickerson, and Jimmy Ba
Keyulu Xu retweeted
Our following 3 papers on graph learning will appear at #ICML2021. Details will follow!
Convergence and implicit acceleration of GNNs
arxiv.org/abs/2105.04550
GraphNorm for accelerating training
arxiv.org/abs/2009.03294
Graph adversarial networks (GAL)
arxiv.org/abs/2009.13504
Our following 3 papers on graph learning will appear at #ICML2021. Details will follow!
Convergence and implicit acceleration of GNNs
arxiv.org/abs/2105.04550
GraphNorm for accelerating training
arxiv.org/abs/2009.03294
Graph adversarial networks (GAL)
arxiv.org/abs/2009.13504
3/3 GNNs are vulnerable to information leakage attacks on social networks etc. We propose GAL with minimax games to protect the sensitive information.
arxiv.org/abs/2009.13504
w/ @LiaoPeiyuan @hanzhao_ml @rsalakhu @StefanieJegelka
2/3 We evaluate and understand what normalization methods work the best for GNNs. Our GraphNorm consistently improves the training and generalization of GNNs.
arxiv.org/abs/2009.03294
w/ @tianle_cai
11
1/3 GNNs are implicitly accelerated by skip connections and more depth. We also show gradient descent training linearized GNNs converge to global minima.
arxiv.org/abs/2105.04550
w/ Kenji Kawaguchi, @mozhi_zhang @StefanieJegelka
Keyulu Xu retweeted
2つの『ポケットモンスター』シリーズ最新作、『ポケットモンスター ブリリアントダイヤモンド・シャイニングパール』と『Pokémon LEGENDS アルセウス』の発売が決定!
シンオウ地方を舞台に、異なる冒険が始まるよ!
pokemon.co.jp/ex/bdsp_legend… #ポケモンBDSP #PokemonLEGENDS
Keyulu Xu retweeted
Today's notes on "How Neural Nets Extrapolate: From Feedforward to GNNs" (arxiv.org/abs/2009.11848) by @KeyuluXu, @StefanieJegelka Ken-ichi Kawarabayashi @SimonShaoleiDu @mozhi_zhang @jingling_li. Read to learn why and when GNNs extrapolate. notion.so/How-Neural-Network…
Our "how neural networks extrapolate" will appear at #ICLR2021 @iclr_conf as an oral presentation.
openreview.net/forum?id=UH-c…
w/ @StefanieJegelka Ken-ichi Kawarabayashi @SimonShaoleiDu @mozhi_zhang @jingling_li
How do neural networks extrapolate, i.e., predict outside the training distribution? We study MLPs and Graph Neural Networks trained by gradient descent, and show how a good representation and architecture can help extrapolation.
arxiv.org/abs/2009.11848
Keyulu Xu retweeted
欲しいいいい!!!
Nintendo Switchに新カラー「マリオレッド×ブルー」登場 1月25日より予約受付開始 nlab.itmedia.co.jp/nl/articl… @itm_nlab
Keyulu Xu retweeted
ICLR's h5-index is 203 and ranks 17 among all scientific publication venues (ahead of NeurIPS, ICCV, ICML).
203 ICLR papers of the last 5 years have gathered more than 203 citations.
Pretty amazing for a conference that started in 2013.
scholar.google.com/citations…
Keyulu Xu retweeted
【拡散希望】前作の発売から半年、なんとこの度、さらにやばい本を姉妹書として上梓しました!!(詳細は画像をご覧下さい)
本当にスーパー面白いので買って下さい!!(ド直球)
『分析者のためのデータ解釈学入門〜データの本質をとらえる技術』Kindle版もすぐ出ます
amazon.co.jp/dp/4802612907/r…
Keyulu Xu retweeted
大統領選などしょうもない話で盛り上がってる間にmultilingual T5が出てたのだけれども、mBERTからの精度の上がり方を見ると、やはり我々NLPerは仕事をするよりも新しい事前学習モデルが出るのを待ちながらあつ森でもやってた方がいいんじゃないかという思いがある。 arxiv.org/abs/2010.11934
Keyulu Xu retweeted
【今日の一曲:11/2】Carla Bruni - Un ange
youtu.be/2pBABLy1Xew
前の前のフランス大統領夫人です
Keyulu Xu retweeted
学習したNNが学習分布外データの予測(外挿)をできるか。MLPは分布外では原点からの方向毎に線形になり、学習目標が非線形な場合は外挿できない。GNNは対象問題が動的計画法の形で解け、適切な非線形がアーキテクチャもしくは特徴に組み込まれている場合、外挿に成功する。arxiv.org/abs/2009.11848
Keyulu Xu retweeted
Ken-ichi Kawarabayashi, Robin Thomas, Paul Wollan: Quickly excluding a non-planar graph arxiv.org/abs/2010.12397 arxiv.org/pdf/2010.12397
Keyulu Xu retweeted
ACL 2021 (@aclmeeting): Preliminary Call for Paper
2021.aclweb.org/calls/papers…
• Anonymity period begins: January 1, 2021
• 𝗦𝘂𝗯𝗺𝗶𝘀𝘀𝗶𝗼𝗻 𝗱𝗲𝗮𝗱𝗹𝗶𝗻𝗲: 𝗙𝗲𝗯𝗿𝘂𝗮𝗿𝘆 𝟭, 𝟮𝟬𝟮𝟭
• Author response period: March 25-31, 2021
• Notification of acceptance: May 5, 2021





























