2020 Machine Learning for Molecules Workshop
Discovering new molecules and materials is a central pillar of human well-being, providing new medicines, securing the world’s food supply via agrochemicals, or delivering new battery or solar panel materials to mitigate climate change. However, the discovery of new molecules for an application can often take up to a decade, with costs spiralling. Machine learning can help to accelerate molecular discovery, which is especially important in light of the Covid19 crisis where drugs/vaccines must be developed to return to normalcy. To reach this goal, it is necessary to have a dialogue between domain experts and machine learning researchers to ensure ML has impact in real world molecular discovery.
In the last decade, progress in machine learning has largely taken place in domains with at least one of two key properties: (1) the input space is continuous, and thus classifiers and generative models are able to smoothly model unseen data that is ‘similar’ to the training distribution, or (2) it is trivial to generate data, such as labelling millions of images or in controlled reinforcement learning settings such as Atari or Go games, where agents can re-play the game millions of times. Unfortunately there are many important problems in chemistry, physics, materials science, and biology that do not share these attractive properties, problems where the data is typically highly structured and discrete and data are typically expensive to gather.
Because of this, there has been a flurry of recent work towards designing machine learning techniques for molecule data. These works have drawn inspiration from and made significant contributions to various areas of machine learning. Recent advances enabled the acceleration of molecular dynamics simulations, increased the efficiency of density based quantum mechanical modeling methods and contributed to accelerate the design of new molecules with optimal properties. This growing field offers unique opportunities for machine learning researchers and practitioners, as it presents a wide spectrum of challenges and open questions, including but not limited to representations of physical systems, physically constrained models, manifold learning, generative modeling of data, interpretability, reliability, model bias, and causality.
The goal of this workshop is to bring together researchers interested in improving applications of machine learning for chemical and physical problems and industry experts with practical experience in pharmaceutical and agricultural development. In a highly interactive format, we will outline the current frontiers and present emerging research directions. We aim to use this workshop as an opportunity to establish a common language between all communities, to actively discuss new research problems, and also to collect datasets by which novel machine learning models can be benchmarked. The program is a collection of invited talks, alongside contributed posters. A panel discussion will provide different perspectives and experiences of influential researchers from both fields and also engage open participant conversation. An expected outcome of this workshop is the interdisciplinary exchange of ideas and initiation of collaboration.
NeurIPS 2020 and Registration
The Machine Learning for Molecules 2020 workshop will be held on December 11 or 12, 2020 as a part of the 34th Annual Conference on Neural Information Processing Systems. Originally planned to be in Vancouver, NeurIPS 2020 and this workshop will take place entirely virtually (online). Please check the main conference website for the latest information. To participate in the workshop, please register for the workshops via the main conference website.
- Benjamin Sanchez-Lengeling (Google Research)
- Frank Noé (FU Berlin)
- Jennifer Listgarten (UC Berkeley)
- Nadine Schneider (Novartis)
- Pat Walters (Relay Therapeutics)
- Rocio Mercado (AstraZeneca)
- Yannick Djoumbou Feunang (Corteva Agriscience)
Call for Papers
The 2020 NeurIPS Workshop on Machine Learning for Molecules calls for high quality contributions on the following topics:
- Applications of ML for molecules (i.e. chemistry, material science, quantum mechanics, biochemistry, biophysics, structural biology, and related areas)
- Graph neural networks
- Representation learning for molecules
- Prediction of molecular properties or bioactivities
- ML for quantum chemistry and molecular dynamics
- Generative models of molecules and de novo design
- ML for protein structures
- ML for reaction data, reaction prediction, and synthesis planning
- Active Learning & Uncertainty Quantification
- Interpretable models
- Benchmarks and analysis of state of the art papers
- Any other topic related to the workshop theme
We call for the submission of breaking edge, high-quality work in progress in the form of extended abstracts or short communication papers, however, finalized projects will also be considered. Work that is presented at the main NeurIPS conference or has been published at other peer-reviewed venues or journals will not be accepted in the workshop. Submissions will be accepted as contributed talks or poster presentations. Accepted final versions will be posted and archived on the workshop website (but do not constitute a proceedings, and can be submitted to other venues/journals).
Submissions should be no longer than 4 pages in NeurIPS style. References do not count towards the page limit. Appendices are discouraged and reviewers will not be required to read beyond the 4 page limit. A broader impact statement is not required.
Papers have to be anonymised for review, and submitted through the EasyChair workshop page.
- Submission Deadline (extended!): 8. October 2020 23:59 Anywhere on earth
- Author notification: 23. October 2020
- Workshop: 11. or 12. December 2020
To be announced.
Organizing Committee and Contact
for questions, please contact firstname.lastname@example.org
- Brooks Paige
- Jennifer Wei
- José Miguel Hernández-Lobato
- Marwin Segler
- Matt Kusner