ML4Materials

from Molecules to Materials

Workshop @ ICLR '23, Fully Virtual, Kigali Rwanda


[Video Recording] [ICLR Page] [OpenReview]


Overview

Many of the world's most crucial challenges, such as access to renewable energy, energy storage, or clean water, are currently fundamentally bottlenecked by materials challenges. The discovery of new materials drives the development of key technologies like solar cells, batteries, and catalysis. Machine learning has significantly impacted the modeling of drug-like molecules and proteins, including the discovery of new antibiotics and the accurate prediction of 3D protein structures. Geometric deep learning methods, in particular, have made tremendous progress in modeling atomic structures and are a promising direction for solving open problems in computational materials science.

While there has been growing interest in materials discovery with machine learning, the specific modeling challenges posed by materials have been largely unknown to the broader community. In particular, compared with the domain of drug-like molecules and proteins, the modeling of materials has the two major challenges outlined below.

First, materials-specific inductive biases are needed to develop successful ML models. For example, materials often don't have a handy representation, like 2D graphs for molecules or sequences for proteins. Moreover, most materials are found in the condensed phase. This means they need to be represented under periodic boundary conditions, introducing challenges to both representation learning and generative models.

Second, there exists a broad range of interesting materials classes, such as inorganic crystals, polymers, catalytic surfaces, nanoporous materials, and more. Each class of materials demands a different approach to represent their structures and new tasks/data sets to enable rapid ML developments.

This workshop aims at bringing together the community to discuss and tackle these two types of challenges. In session A, we will feature speakers to discuss the latest progress in developing ML models for materials focusing on algorithmic challenges, covering topics like geometric deep learning and generative models. In particular, what can we learn from the more developed field of ML for molecules and proteins, and where might challenges differ and opportunities for novel developments lie? In session B, we will feature speakers to discuss unique challenges for each sub-field of materials design and how to define meaningful tasks that are relevant to the domain, covering areas including inorganic materials, polymers, nanoporous materials, and catalysis. More specifically, what are the key materials design problems that ML can help tackle?


Call for papers


We encourage the community to contribute 4-page extended abstracts to our workshop. Example topics include (but not limited to):

All submissions are required to use the ICLR style file. References and appendices do not count towards the page limit, but reviewers will not be required to read beyond the first 4 pages. All accepted papers will be non-archival. Papers have to be anonymized for review, and submitted through the OpenReview workshop page.


Important dates



Schedule (in US EST)


09:00 - 09:10         Opening Remark

09:10 - 09:40         Invited Talk, Boris Kozinsky
                                       

09:40 - 10:10         Invited Talk, Marivi Fernandez-Serra
                                        Machine learning approaches to improve the exchange and correlation functional in Density functional Theory

10:10 - 10:30         Break

10:30 - 11:00         Invited Talk, Tess Smidt
                                        Harnessing the properties of equivariant neural networks to understand and design materials

11:00 - 11:30         Invited Talk, Andrew Ferguson
                                        Machine learning-guided directed evolution of functional proteins

11:30 - 12:00         Spotlight Talks
                                        JAX-XC: Exchange Correlation Functionals Library in Jax
                                        Crystal Structure Prediction by Joint Equivariant Diffusion on Lattices and Fractional Coordinates
                                        Cross-Quality Few-Shot Transfer for Alloy Yield Strength Prediction: A New Material Science Benchmark and ...

12:00 - 13:00         Poster Session 1

13:00 - 13:30         Break

13:30 - 14:00         Invited Talk, Yousung Jung
                                        Machine learning to generate molecules and materials and their synthesis predictions

14:00 - 14:30         Invited Talk, Rafael Gomez-Bombarelli
                                       

14:30 - 14:50         Break

14:50 - 15:20         Invited Talk, Shyue Ping Ong
                                        A potential of everything

15:20 - 15:50         Invited Talk, Zachary Ulissi
                                        Open datasets/models in catalysis: recent progress their use to massively accelerate adsorption energy workflows

15:50 - 16:00         Break

16:00 - 17:00         Panel Discussion
                                        Boris Kozinsky · Tess Smidt · Rafael Gomez-Bombarelli · Marivi Fernandez-Serra · Zachary Ulissi · Shyue Ping Ong ·
                                        Yousung Jung · Andrew Ferguson

17:00 - 18:00         Poster Session 2

18:00 - 18:10         Closing Remark

-->

Invited Speakers



Organizing Committee



Advising Committee



Program Committee, Reviewers


We are now actively looking for reviewers/PCs, if you are interested, please fill out this Google form!


Accepted Papers


Machine learning-assisted close-set X-ray diffraction phase identification of transition metals
Maksim Zhdanov; Andrey Zhdanov

Transfer Learning with Diffusion Model for Polymer Property Prediction
Gang Liu; Meng Jiang

Fragment-based Multi-view Molecular Contrastive Learning
Seojin Kim; Jaehyun Nam; Junsu Kim; Hankook Lee; Sungsoo Ahn; Jinwoo Shin

Learning single-step retrosynthesis with pseudo-reactions
Shuan Chen; Yousung Jung

CrysGNN : Distilling pre-trained knowledge to enhance property prediction for crystalline materials.
KISHALAY DAS; Bidisha Samanta; Pawan Goyal; Seung-Cheol Lee; Satadeep Bhattacharjee; Niloy Ganguly

3D Graph Conditional Distributions via Semi-Equivariant Continuous Normalizing Flows
Eyal Rozenberg; Ehud Rivlin; Daniel Freedman

Cooperative data-driven modeling: continual learning of different material behavior
Aleksandr Dekhovich; Ozgur Taylan Turan; Jiaxiang Yi; Miguel Anibal Bessa

Crystal Structure Prediction by Joint Equivariant Diffusion on Lattices and Fractional Coordinates[Oral]
Rui Jiao; Wenbing Huang; Peijia Lin; Jiaqi Han; Pin Chen; Yutong Lu; Yang Liu

Compositional and elemental descriptors for perovskite materials
Jiri Hostas; Maicon Pierre Lourenço; John Garcia; Hatef Shahmohamadi; Alain Tchagang; Karthik Shankar; Venkataraman Thangadurai; Dennis R. Salahub

Latent Conservative Objective Models for Offline Data-Driven Crystal Structure Prediction
Han Qi; Stefano Rando; Xinyang Geng; Iku Ohama; Aviral Kumar; Sergey Levine

Cross-Quality Few-Shot Transfer for Alloy Yield Strength Prediction: A New Material Science Benchmark and An Integrated Optimization Framework[Oral]
Xuxi Chen; Tianlong Chen; Everardo Yeriel Olivares; Kate Elder; Scott McCall; Aurelien Perron; Joseph McKeown; Bhavya Kailkhura; Zhangyang Wang; Brian Gallagher

JAX-XC: Exchange Correlation Functionals Library in Jax [Oral]
Kunhao Zheng; Min Lin

Predicting Density of States via Multi-modal Transformer
Namkyeong Lee; Heewoong Noh; Sungwon Kim; Dongmin Hyun; Gyoung S. Na; Chanyoung Park

Graph-informed simulation-based inference for models of active matter
Namid Stillman; Silke Henkes; Roberto Mayor; Gilles Louppe

Matbench Discovery - Can machine learning identify stable crystals?
Janosh Riebesell; Rhys Goodall; Anubhav Jain; Kristin Persson; Alpha Lee

Machine Learning for XRD Spectra Interpretation in High-Throughput Material Science
Hilary Egan; Davi Marcelo Febba; Andriy Zakutayev

In-situ Anomaly Detection in Additive Manufacturing with Graph Neural Networks
Sebastian Larsen; Paul A. Hooper

Controlling Dynamic Spatial Light Modulators using Equivariant Neural Networks
Sumukh Vasisht Shankar; Darrel D'Souza; Jonathan P Singer; Robin Walters

Designing Nonlinear Photonic Crystals for High-Dimensional Quantum State Engineering
Eyal Rozenberg; Aviv Karnieli; Ofir Yesharim; Joshua Foley-Comer; Sivan Trajtenberg-Mills; Sarika Mishra; Shashi Prabhakar; Ravindra Pratap Singh; Daniel Freedman; Alex M. Bronstein; Ady Arie

Behavioral Cloning for Crystal Design
Prashant Govindarajan; Santiago Miret; Jarrid Rector-Brooks; Mariano Phielipp; Janarthanan Rajendran; Sarath Chandar

A framework for fully autonomous design of materials via multiobjective optimization and active learning: challenges and next steps
Tyler H Chang; Jakob R Elias; Stefan M. Wild; Santanu Chaudhuri; Joseph A. Libera

MatKB: Semantic Search for Polycrystalline Materials Synthesis Procedures
Xianjun Yang; Stephen Wilson; Linda Petzold

Forward and Inverse design of high $T_C$ superconductors with DFT and deep learning
Daniel Wines; Kevin F Garrity; Tian Xie; Kamal Choudhary

Expanding the Extrapolation Limits of Neural Network Force Fields using Physics-Based Data Augmentation
Yuliia Orlova; Gavin Keith Ridley; Frederick Zhao; Rafael Gomez-Bombarelli

SimuStruct: Simulated Structural Plate with Holes Dataset with Machine Learning Applications
Bruno Alves Ribeiro; Joao Alves Ribeiro; Faez Ahmed; Hugo Penedones; Jorge Belinha; Luís Sarmento; Miguel Bessa; Sérgio Tavares

Constructing and Compressing Global Moment Descriptors from Local Atomic Environments
Vahe Gharakhanyan; Max Shirokawa Aalto; Aminah Alsoulah; Nongnuch Artrith; Alexander Urban


Contact


If you have any questions, you can reach out to organizers at ml4materials@googlegroups.com


Sponsor



The webpage template is by the courtesy of CVPR 2020 Workshop on Adversarial Machine Learning in Computer Vision.