Scientific Machine Learning Mini-Course
The mini-course organized by Dilip Krishnamurthy and Venkat Viswanathan is a short seminar series (12+ weeks starting Oct 1, 2020) with the goal of cross-pollinating ideas between the various emerging methods at the intersection of physics and machine learning.
Seminar 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.
Seminar Time: Thursdays 8:30 pm to 10 pm Eastern Time
Speakers and Topics
Physics-Regularized ML:
- Oct 1: Dilip Krishnamurthy, Carnegie Mellon University PhD
Physics-Informed Neural Networks
Session Chair: Bharath Ramsundar, Stanford PhD - 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: 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: Jan Hermann
, Freie Universität Berlin & Humboldt University of Berlin Physics PhD
Deep neural network solution of the electronic Schrödinger equation - Nov 5: Tess Smidt, Lawrence Berkeley Laboratory & University of California Berkeley PhD
Euclidean Symmetry Equivariant Neural Networks
Session Chair: Prof. Risi Kondor, University of Chicago
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 perspective:
- Joel Moxley, Breakthrough Energy
- Lisha Li, Rosebud AI
- James Hardiman, DCVC (Data Collective)
Industry perspective:
- Shailendra Kaushik, General Motors
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: Alec Bills, Carnegie Mellon University PhD
Universal Ordinary Differential Equations for Battery Performance and Reliability Predictions - Nov 26: Jesse Bettencourt
, University of Toronto PhD
Neural Ordinary Differential Equations - TBD: Shaojie Bai
, Carnegie Mellon University PhD
Deep Equilibrium Models - TBD: Edwin Khoo, Institute for Infocomm Research & Massachusetts Institute of Technology PhD
Physics-Aware Machine Learning for Battery Performance Predictions - TBD: Miles Cranmer, Princeton Astrophysics PhD
Lagrangian Neutral Networks, Graph Nets and Symbolic regression - Dec 3: Alok Warey, General Motors & University of Texas at Austin PhD
Deep Learning for Vehicle Systems - Dec 10: Linfeng Zhang, Princeton PhD
Integrating Machine Learning with Physics-Based Modeling - Dec 17: Dhairya Gandhi, Julia Computing
Scientific Machine Learning Methods and Tools (DiffEqFlux.jl) - Jan 7: Varun Shankar, Carnegie Mellon University
Physics-Constrained Machine Learning for Fluid Flow Fields - Jan 14: Rachel Kurchin, Carnegie Mellon University & Massachusetts Institute of Technology PhD
Physics-Guided Convolutional Neural Networks - Jan 21: Gurtej Kanwar, Massachusetts Institute of Technology PhD
Gauge-Invariant Machine-Learned Flow-Based Sampling Algorithms
Overleaf Template for Methodology Walk-Through
Colab Notebook Format for Code Walk-Through
How To Join
Add Webinar Calendar
Zoom Link
Webinar ID: 992 4479 8052
Passcode: 919401
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 dkrishn1[at]andrew.cmu.edu and venkvis[at]cmu.edu