George Papamakarios

@gpapamak

Research Scientist @GoogleDeepMind, working on Gemini pretraining

London, England

gpapamak.github.io

Joined December 2017

394 Following

2,555 Followers

Replies Reposts Media

George Papamakarios
@gpapamak
May 26, 2023
Happy to share new work in #AI4Science Gibbs free energies via isobaric-isothermal flows arxiv.org/abs/2305.13233 In previous work we proposed a flow model that generates crystals at constant volume – this one handles the (harder) case of varying volume at constant pressure
1
George Papamakarios
@gpapamak
Apr 26, 2023
📢 New paper on diffusion models for simulation-based inference, to appear at #ICML2023 arxiv.org/abs/2209.14249 tl;dr: Diffusion models are great for SBI: using a trick akin to classifier-free guidance, they let you do inference with more observations than you trained on 🙂
3
George Papamakarios
@gpapamak
Feb 13, 2023
New paper: Equivariant MuZero arxiv.org/abs/2302.04798 A version of MuZero that incorporates environment symmetries by design. Great internship work by @andreeadeac22 👏👏👏
Andreea Deac and Theo Weber
1
George Papamakarios
@gpapamak
Dec 8, 2022
Happy to share our latest work: "A Generalist Neural Algorithmic Learner" A single graph neural network that learns to execute 30 different algorithms: sorting, searching, string matching, etc Challenging multi-task & generalization problem where inductive biases are important
Google DeepMind
@GoogleDeepMind
Dec 8, 2022
Introducing a generalist neural algorithmic learner, capable of carrying out 30 different reasoning tasks, with a 𝘴𝘪𝘯𝘨𝘭𝘦 graph network. These include: 🔵 Sorting 🔵 Shortest paths 🔵 String matching 🔵 Convex hull finding And more: dpmd.ai/3FC1FqA
1
George Papamakarios
@gpapamak
Apr 22, 2022
Code release: We have open-sourced #JAX code and released pre-trained models for "Normalizing flows for atomic solids". github.com/deepmind/flows…
George Papamakarios
@gpapamak
Nov 17, 2021
Sharing new work in the intersection of machine learning & physics: Normalizing flows for atomic solids arxiv.org/abs/2111.08696 Shows that flows can model crystalline solids well enough to estimate free energy accurately, even with the simplest importance-weighted estimator.

George Papamakarios

George Papamakarios’s posts