Scientific Machine Learning Webinar Series

The SciML webinar has moved to the University of Michigan's MICDE. Find the website, calendar, Zoom link and talks at https://micde.umich.edu/news-events/sciml-webinar-series/ . If you were on our mailing list, look out for an email to consent being added to the new SciML webinar mailing list.

This webinar series and panel events are organized by Keith Phuthi, Varun Shankar and Venkat Viswanathan with the goal of cross-pollinating ideas between the various emerging methods at the intersection of physics and machine learning.

Webinar Format: Presenters can use the opportunity to showcase a paper or two with an explicit focus on the methodology and approach. Duration: 40 minutes of methodology + 20 minutes of implementation (code) walk-through + 20 minutes of questions. Invited session chairs will guide the discussion along with offering their perspective on the field. The Q&A session is typically very interactive with a small group of enthusiastic audience.

Seminar Time: Thursdays 11 am to 12:30 pm Eastern Time

Suggest speakers:

Form

Past Webinars

Spring 2023:

• March 23: Colabfit Team with Stefano Martiniani (NYU), Eric Fuemmeler (UMN) and Amit Gupta (UMN)
  Tentative Title: ML Interatomic Potentials development within the OpenKIM/ColabFit frameworks.
  Session Chair:
• March 30: Xiaoyu Xie (Northwestern)
  Tentative Title: Data-driven discovery of dimensionless numbers and governing laws from scarce measurements
  Session Chair: Dr. Youngsoo Choi (Lawrence Livermore National Lab)
• April 6: Leon Gerard and Michael Scherbela (University of Vienna)
  Tentative Title: Neural Network Wavefunctions with Variational Monte Carlo
  Session Chair: Weiluo Ren (Bytedance)
• April 13: Edwin Garcia (Purdue)
  Title: Machine Learning of Phase Diagrams
  Session Chair: Ursula Kattner (NIST)
• April 20: Shengjie Luo (Peking University)
  Tentative Title: One Transformer Can Understand Both 2D & 3D Molecular Data
  Session Chair: Payel Das (IBM)
• April 27: Misaki Ozawa (Université de Grenoble Alpes)
  Tentative Title: Renormalization Group Approach for Machine Learning Probability Distributions
  Session Chair: Bruno Loureiro (École Normale Supérieure)
• May 11: Saro Passaro (Meta AI)
  Tentative Title: Reducing SO(3) Convolutions to SO(2) for Efficient Equivariant GNNs
  Session Chair: Josh Rackers (Genentech)
• May 18: Chenru Duan (MIT)
  Tentative Title: A transferable recommender approach for selecting the best density functional approximations in chemical discovery
  Session Chair: Matthew Welborn (Entos)
• May 25: Marylou Gabrié (École Polytechnique)
  Tentative Title: Adaptive Monte Carlo augmented with normalizing flows
  Session Chair: Tony Lelièvre (CERMICS)

Fall 2022:

• September 22: Batatia Ilyes (ENS Paris Saclay)
  MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields
  Session Chair: Mario Geiger (Massachusetts Institute of Technology)
• September 29: Alby Musaelian and Simon Batzner (Harvard University)
  Learning Local Equivariant Representations for Large-Scale Atomistic Dynamics
  Session Chair: Bharath Ramsundar (Deep Forest Sciences)
• October 6: Yongji Wang (Princeton University)
  Solving 3-D Euler via PINNs
  Session Chair: Lu Lu (University of Pennsylvania)
• October 20: Jan Weinreich (University of Vienna)
  Ab-initio machine learning of phase space averages
  Session Chair: Stephan Heinen (University of Vienna)
• October 27: Darshil Doshi (University of Maryland College Park)
  Critical Initialization of Deep Neural Networks using Jacobians
  Session Chair: Tankut Can (Institute for Advanced Study)
• November 3: Gert-Jan Both (Université de Paris)
  Fully Differentiable Model Discovery
  Session Chair: Christian Mueller (Helmholtz Center Munich)
• December 1: Yu Shelby Xie (Harvard University)
  FLARE: Many-body Bayesian force field and uncertainty-aware molecular dynamics from Bayesian active learning
  Session Chair: Chris Paolucci (University of Virginia)

Generative Models:

• July 14: Youzhi Luo (Texas A&M)
  An Autoregressive Flow Model for 3D Molecular Geometry Generation
  Session Chair: Jian Tang (Mila-Quebec AI Institute)
• July 28: Yi Liu (Texas A&M)
  Complete and Efficient Graph AI Techniques for Molecular Sciences
  Session Chair: Wengong Jin (Broad Institute)
• August 4: Yuelin Wang and Kai Yi (Shanghai Jiao Tong University)
  ACMP: Allen-Cahn Message Passing for Graph Neural Networks with Particle Phase Transition
  Session Chair: Jia Zhao (Utah State University)
• August 18: Niklas Gebauer (TU Berlin)
  Inverse design of 3d molecular structures with conditional generative neural networks
  Session Chair: Gregor Simm (Microsoft Research)

Differentiable Physics:

• February 24: Giuseppe Romano (MIT)
  Differentiable Phonon Simulations to Optimize Thermal Transport in Nanostructures
  Session Chair: Ján Drgoňa (PNNL)
• March 3: Patrick Kidger (University of Oxford)
  On Neural Differential Equations
  Session Chair: David Duvenaud (University of Toronto)
• March 10: Joe Greener (MRC Laboratory of Molecular Biology)
  Differentiable molecular simulation can learn all the parameters in a coarse-grained force field for proteins
  Session Chair: Sam Schoenholz (Google Brain)
• March 17: Rafael Gomez-Bombarelli (MIT)
  Differentiable Uncertainty
  Session Chair: Olexandr Isayev (CMU)
• March 24: Desmond Zhong (Siemens Technology)
  Extending Lagrangian and Hamiltonian Neural Networks with Differentiable Contact Models
  Session Chair: Rachel Kurchin (CMU)
• March 31: Taylor Howell & Simon Le Cleac'h (Stanford)
  DOJO - differentiable rigid-body-dynamics
  Session Chair: Vikas Sindhwani (Google Brain)
• April 21: Akshay Agrawal
  Networks with Differentiable Contact Models
  Session Chair: Zac Manchester (CMU)

Symmetries, Physical Systems and Machine Learning:

• October 14: Rui Wang (UCSD) and Robin Walters (Northeastern)
  Incorporating Symmetry for Improved Generalization in Dynamics Prediction
  Session Chair: Soledad Villar (Johns Hopkins University)
• October 28: Johannes Brandstetter (Johannes Kepler University Linz)
  Geometric and Physical Quantities Improve E(3) Equivariant Message Passing
  Session Chair: Stephan Günnemann (Technical University of Munich)
• November 4: Yu Wang (MIT)
  Geometric Operators for Shape Analysis
  Session Chair: Rana Hanocka (University of Chicago)
• November 18: Pim de Haan (Qualcomm)
  Natural Message Passing
• December 2: Carlos Esteves (Google)
  Towards Efficient Spherical NNs
  Session Chair: Manzil Zaheer (Google Research)
• January 6: Ziming Liu (MIT)
  Machine Learning Symmetries for Conservation Laws
  Session Chair: Sam Vinko (University of Oxford)
• January 13: Zhuoran Qiao (Caltech)
  Geometric Learning for Quantum-Chemistry-Informed Representations
  Session Chair: Peetak Mitra (PARC)
• January 20: Ferran Alet (MIT)
  Learning to Encode and Discover Physics-Based Inductive Biases
  Session Chair: Bharath Ramsundar (Deep Forest Sciences)

Machine Learning Potentials and Force Fields for Materials Chemistry:

• August 19: Pratyush Tiwary, University of Maryland
  From Atoms to Mechanisms with State Predictive Information Bottleneck and Denoising Diffusion Proabilistic Models
  Session Chair: Matthias Rupp, University of Konstanz
• August 26: Sam Schoenholz , Google Brain
  JAX-MD: A Framework for Differentiable Molecular Dynamics
  Session Chair: Michael Brenner, Harvard University
• September 2: Linfeng Zhang, Princeton University
  Learning-Assisted Molecular Modeling: From Methodology Development ot Engineering Effort
  Session Chair: Tess Smidt, Massachusetts Institute of Technology
• September 9: Simon Batzner, Harvard University
  Neural Equivariant Interatomic Potentials
  Session Chair: Matti Hellström, Software for Chemistry and Materials
• September 16: Alice Allen and Dávid Péter Kóvacs, University of Cambridge
  Linear Body-Ordered Molecular Force Fields
  Session Chair: Albert Bartók-Pártay, University of Warwick
• September 23: Andrea Grisafi, École Polytechnique Fédérale de Lausanne
  Symmetry-Adapted and Long-Range Representations in Atomic-scale ML
  Session Chair: Kieron Burke, University of California Irvine
• September 30: Muhammad Firmansyah Kasim and Sam Vinko, University of Oxford
  Differentiable Quantum Chemistry
  Session Chair: Ekin Dogus Cubuk, Google Brain

Machine Learning meets Information Theory and Statistical Mechanics:

• July 8: Alex Alemi, Google Research
  Machine Learning and Thermodynamics
  Session Chair: Max Welling, University of Amsterdam
• July 15: Pratik Chaudhri, University of Pennsylvania
  (Towards the) Foundations of Small Data
  Session Chair: Karthik Duraisamy, University of Michigan
• July 22: Sho Yaida, Facebook
  Model Complexity from Macroscopic Perspective
  Session Chair: Jascha Sohl Dickstein, Google Brain
• July 29: Yasaman Bahri, Google Brain
  Dynamics and Phase Transitions in Wide, Deep Neural Networks
  Session Chair: Surya Ganguli, Stanford
• Aug 5: Elena Agliari, Sapienza University of Rome
  Learning From Storing
  Session Chair: Jascha Shol-Dickstein, Google Brain

Machine Learning in Fluid Dynamics:

• May 20: Filipe de Avila Belbute-Peres, Carnegie Mellon University
  Combining PDE Solvers and Graph Neural Networks for Fluid Flow Prediction
  Session Chair: Karthik Kashinath, Berkeley Lab
• May 27: Joseph Bakarji, University of Washington
  Data-driven Discovery of Differential Equations for Complex Models in Fluids
  Session Chair: Julia Ling, Alphabet
• June 3: Pedro M. Milani, Exponent
  ML Approaches to Learning Turbulent Mixing in Film Cooling Flows
  Session Chair: Gavin Portwood, LLNL
• June 10: Zongyi Li, Caltech
  Neural Operator: Learning Maps Between Function Spaces
  Session Chair: Sanjay Choudhry, NVIDIA
• June 17: Ameya D. Jagtap, Brown University
  A Generalized Space-Time Domain based Extended PINN for PDEs: Method and Implementation
  Session Chair: Rose Yu, UCSD
• June 24: Dmitrii Kochkov, Google
  Machine Learning Accelerated Computational Fluid Dynamics
  Session Chair: Themistoklis Sapsis, MIT
• July 1: Pierre Baque, Neural Concept
  Session Chair: Andrea Panizza, Baker Hughes

VC Panel on Quantum Computing (May 11th, Tuesday at 2 pm Pacific Time):

• Moderator: Sridhar Tayur, Carnegie Mellon University
• Carly Anderson, Prime Movers Lab
• Chris Boshuizen, DCVC
• Russ Wilcox, Pillar VC

Quantum Machine Learning:

The organizers would like to thank Jarrod McLean (Google) and Zlatko K. Minev (IBM) for suggestions of speakers and session chairs.

• April 15: Hsin-Yuan (Robert) Huang , Caltech
  Characterizing Quantum Advantage in Machine Learning
  Session Chair: Kristan Temme, Institute for Quantum Information and Matter
• April 22: Ian Convy , UC Berkeley
  Session Chair: Miles Stoudenmire, Flatiron Institute
• April 29: Andrea Skolik, Volkswagen Data Lab and Leiden University
  Session Chair: Maria Schuld, Xanadu and University of KwaZulu-Natal
• May 6: Alba Cervera Lierta, University of Toronto
  Session Chair: Glen Evenbly, Georgia Institute of Technology
• May 13: Michael Broughton, Google
  Session Chair: Max Radin, Zapata Computing

Deployment of ML in the Industry:

• Mar 11: Melanie Senn and Alina Negoita, Innovation Center California of Volkswagen Group of America
  High-Throughput Screening Framework for Battery Materials Design
  Session Chair: Shailendra Kaushik, General Motors
• Mar 18: Austin Sendek, Aionics, Inc.
  Aionics: Harnessing ML to supercharge battery discovery, design, and deployment in industry
  Session Chair: Venkat Viswanathan, Carnegie Mellon University
• Mar 25: Payel Das, IBM Thomas J Watson Research Center
  Trustworthiness in AI for Accelerating Discovery
  Session Chair: Isidoros Doxas, Northrop Grumman Mission Systems
• Apr 1: Aniruddha Mukhopadhyay, ANSYS, Inc.
  Shifting Landscape in ML-Driven Product Engineering
  Session Chair: Vivek Singh, NVIDIA
• Apr 8: Keith Task, BASF
  Data Science for Chemicals and Materials Development at BASF
  Session Chair: Amra Peles, Pacific Northwest National Laboratory

Molecular ML for Drug Discovery:

• Feb 11: Wengong Jin, Massachusetts Institute of Technology
  Graph Neural Networks and Generative Models for Drug Discovery
  Session Chair: Alex Wiltschko, Google Research
• Feb 18: Bharath Ramsundar and Seyone Chithrananda
  ChemBERTa: Large-Scale Self-Supervised Pretraining for Molecular Property Prediction
  Session Chair: Tom Miller, Caltech and Entos, Inc.
• Feb 25: Dominik Lemm
  Energy-Free Machine Learning Predictions of Ab Initio Structures
• March 4: Mario Krenn
  Robust Molecular String Representation for Molecular Machine Learning
  Session Chair: Olexandr Isayev, Carnegie Mellon University

Panel Discussion on Open Challenges in ML:

This session is focused on discussing challenges and technological bottlenecks at the intersection of machine learning and science/engineering. Industry leaders at original equipment manufacturers (OEMs) and venture capitalists (VCs) will provide their perspective and directions for research and development. We anticipate that this session will facilitate effective TT & O (Tech. Transfer and Outreach).

VC Panel (March 23rd):

• Joel Moxley, Breakthrough Energy Ventures
• James Hardiman, DCVC (Data Collective)
• Mark Cupta, Prelude Ventures
• Bilal Zuberi, Lux Capital

Physics-Regularized ML:

• Oct 1: Dilip Krishnamurthy, Carnegie Mellon University PhD
  Physics-Informed Neural Networks
  Session Chair: Bharath Ramsundar, DeepChem
• Oct 8: Xiaowei Jia, Asst. Prof. @ University of Pittsburgh & University of Minnesota PhD
  Physics-Guided Machine Learning for Scientific Discovery
  Session Chair: Jason Koeller, Citrine Informatics

ML Obeying Physical Symmetries:

• Oct 15 (note schedule change): Bingqing Cheng, Trinity College Cambridge & EPFL PhD
  Machine-Learning Potentials: What Works and What Doesn’t
  Session Chair: Prof. Frank Noe, Freie Universität Berlin
• Oct 22: Bharath Ramsundar, Creator of DeepChem & Stanford CS PhD
  Physical Theories and Differentiable Programs
  Session Chair: Prof. Venkat Viswanathan, Carnegie Mellon University
• Oct 29 (note schedule change): Jan Hermann , Freie Universität Berlin & Humboldt University of Berlin Physics PhD
  Deep neural network solution of the electronic Schrödinger equation
  Session Chair: Prof. Giuseppe Carleo, École polytechnique fédérale de Lausanne (EPFL)
• Nov 5 (note schedule change): Tess Smidt and Mario Geiger, Lawrence Berkeley Laboratory & École polytechnique fédérale de Lausanne, respectively
  Neural Networks With Euclidean Symmetry for Physical Sciences and e3nn: A Modular PyTorch Framework for Euclidean Neural Networks
  Session Chair: Prof. Risi Kondor, University of Chicago

ML-Embedded Physical Models:

• Nov 12: Li Li, Google Accelerated Science & UC Irvine PhD
  Kohn-Sham equations as regularizer: building prior knowledge into machine-learned physics
  Session Chair: Prof. Vikram Gavini, University of Michigan
• Nov 19: Christopher Rackauckas and Alec Bills, Massachusetts Institute of Technology & Pumas-AI, and Carnegie Mellon University, respectively
  Universal Ordinary Differential Equations and Its Application to an Engineering Challenge
  Session Chair: Prof. Srinivas (Sai) Ravela, Massachusetts Institute of Technology
• Dec 3: Alok Warey, General Motors & University of Texas at Austin PhD
  Deep Learning for Vehicle Systems
  Session Chair: Aniruddha Mukhopadhyay, ANSYS, Inc.
• Dec 10: Jesse Bettencourt , University of Toronto PhD
  Neural Ordinary Differential Equations
  Session Chair: Prof. Zico Kolter, Carnegie Mellon University
• Jan 14: Miles Cranmer, Princeton Astrophysics PhD
  Time Symmetries and Neurosymbolic Learning for Dynamical Systems
  Session Chair: Prof. Phiala Shanahan, Massachusetts Institute of Technology
• Jan 21: Shaojie Bai, Carnegie Mellon University PhD
  Deep Equilibrium Models
  Session Chair: Stephan Hoyer, Google Research
• Jan 28: Gurtej Kanwar, Massachusetts Institute of Technology PhD
  Gauge Equivariant Normalizing Flows for Lattice Field Theory
  Session Chair: Lena Funcke, Perimeter Institute for Theoretical Physics
• Feb 4: Evan Feinberg, Genesis Therapeutics, Inc.
  Machine Learning and Molecular Simulation Based Methods for Therapeutics
  Session Chair: Amir Barati Farimani, Carnegie Mellon University

Resources

The seminar series is supported by the ARPA-E DIFFERENTIATE program and the Carnegie Mellon Presidential Fellowship.

Conception

A panel discussion on the topic Machine Learning Based Approaches to Accelerate Energy Materials Discovery and Optimization crystallized important considerations while applying machine learning methods to limited-data engineering applications. Often it's useful to synergistally stack the ML models to the extent possible with the known physics of the problem for effective learning even in low-data regimes.

Questions?

Email the organizers at mkphuthi[at]cmu.edu and venkvis[at]cmu.edu

Video Recordings

A Google Drive link to the Video recordings are available on request on the slack or by emailing the organizers.