NSF Workshop on Data-driven Modeling and Prediction of Rare and Extreme Events
Registration for this workshop has now closed. If you have already submitted an application, you will hear back by September 20.
Overview
This workshop will convene experts in rare and extreme event detection and characterization representing a broad range of application domains and disciplines, including statistics, machine learning, applied mathematics, space weather, materials science, and climate modeling. The goal is for these experts with complementary backgrounds to identify key challenges and opportunities, with an emphasis on methodologies that may be leveraged across domains. The focus on data-driven methods encompasses recent efforts in machine learning, including physics-informed machine learning and generative models, and how such tools may advance rare and extreme event forecasting. The agenda will also include physics-driven approaches, including simulations, both as a source of fundamental insights into the modeling of rare events and as a mechanism for generating data to complement real-world data used to train data-driven models. This two-day workshop will be held at the University of Chicago on November 21-22, 2024. It will feature lectures from experts across the spectrum of disciplines listed above, panel discussions, poster sessions, and lightning talks.
Workshop theme and goals
The detection, characterization, and forecasting of rare events play critical roles in many science and engineering settings, including non-destructive testing of materials, space weather and climate forecasting, tipping points in ecological systems, and cascading series of failures in the power grid. In these and other applications, there is growing interest in using data-driven methods (e.g., machine learning) to increase accuracy, efficiency, and robustness. However, many off-the-shelf machine learning tools are ill-suited to the modeling and prediction of rare events because they are rare. For instance, a classifier may misclassify all rare events and still have high average accuracy, and generative models may fail to generate samples in the tails of distributions underlying training data. Such failures may not impact standard performance metrics appreciably, but have a heavy impact on downstream scientific inquiry and engineering efforts. In parallel, sophisticated simulation systems provide unique opportunities for methods of the interface of simulation and machine learning. A key objective of the workshop is to develop an understanding of how we can ensure data-driven methods are sensitive to and even exploit rare and extreme events, leverage physical models and simulators, and facilitate the prediction of future incipient failures (e.g., aircraft vulnerabilities or space weather impacts).
The goal of this workshop is to identify existing tools, open problems, and research opportunities for forecasting and characterizing rare extreme events using data-driven methods. Our focus with this workshop is on mechanisms that detect, predict, and characterize rare and extreme events.
Agenda
Thursday, November 21, 2024
Welcome
Workshop Overview, Themes, and Goals
Foundational methods
Talk title: Predicting grey swans from time series data with machine learning
Break
Turbulent dynamic systems
Talk title: Quantification of extreme events in engineering and geophysical turbulence
Lightning talks
Simon Frost: Rare events in vector-borne disease surveillance
Daniel Schwalbe-Koda: Model-free Quantification of Outliers in Atomistic Simulations and Machine Learning with Information Theory
Yao Xie: Flow-based distributionally robust optimization
John Wertz: Deep learning applied to electrical impedance tomography on structural materials
Mihai Anitescu: Issues in Modeling and Computing with Extremes
Azim Ahmadzadeh: Your Performance Measure Is Not What You Think It Is!
Ashwin Renganathan: Sample efficient estimation of and optimization with rare events
Lunch and poster session
Ted Cary: Physics-informed machine learning for nondestructive evaluation
Nisha Chandramoorthy: TBA
Rui Ding: Hunting for Rare High-Performance Electrocatalysts: A Multi-Stage ML Approach to Discover Extreme Performers in Acidic Oxygen Evolution
Joel Harley: Self-Calibrating Unsupervised Learning for Long-Term Structural Health Monitoring in Dynamic Environments
Olawale Ikuyajolu: Machine Learning Emulation of Non-linear Wave-wave Interaction Source Terms in a Wave Model
Ruoxi Jiang: TBA
Ling-Wei Kong: Anticipating Dynamical Transitions by Machine Learning
Xingjie Li: Nyström-type inference-based schemes adaptive to large time-stepping for long-term simulation
Akwum Onwunta: Tensor train solution to optimization problems constrained by differential equations under uncertainty
Daniel Schwalbe-Koda: Model-free Quantification of Outliers in Atomistic Simulations and Machine Learning with Information Theory
Shashank Sule: Learning collective variables for accurate representations molecular dynamics and rare events
Shanyin Tong: Large deviation theory-based adaptive importance sampling for rare events in high dimensions
Osei Tweneboah: TBA
Yinan Wang: Advanced Machine Learning for Anomaly Detection
John Wertz: Deep learning applied to electrical impedance tomography on structural materials
Alexander Wikner: TBA
Hao Yan: Personalized Tensor Decomposition: Decoupling Commonality and Peculiarity
Space weather
Talk title: Fifty-for-One Isn’t Fair: Power-Ups to Level the Field in Data-Driven Solar Flare Forecasting
Materials science
Talk title: Data-Driven Approaches to Establish Process-Porosity-Fatigue Relationships in Metal Additive Manufacturing
Break
Panel discussion: Perspectives on national priorities
Climate modeling and forecasting
Talk title: Improving Distributional Learning for Dynamic Systems
Reception
Friday, November 22, 2024
Foundational methods
Talk title: Characterizing rare events in epidemic models
Materials science
Talk title: Nondestructive Evaluation at NASA: Finding the Needle Without Disturbing the Haystack
Lightning talks
Nicholas Boffi: Elucidating the nonequilibrium dynamics of flocking with deep learning
Paul Schrader: The Topological Data Analysis Machine Learning Algorithm (TDAML) and its Emerging Implementations
Yuehaw Khoo: Re-Anchoring Quantum Monte-Carlo with Tensor-Train Sketching
Nan Chen: Integrating Data Assimilation with Machine Learning for Advancing Extreme Event Modeling
Abani Patra: Using a Large Deviations based Anomaly Detection Model for Improving neural network training
Whitney Huang: Estimating Storm Surge Risk: A Physical-Statistical Approach
Alexander Rodriguez: Real-time Disease Outbreak Response by Bridging AI, Data, and Scientific Models
Shanyin Tong: Large deviation for rare event studies: bridging probability theory with PDE-constrained optimization
Break
Space weather
Talk title: Rare Events in Space Weather
Climate modeling and forecasting
Talk title: Transformed-Linear Methods for Extremes and Fire Season Attribution
Lunch with poster session
Mahshid Ahmadian: A New Methodology for Spatiotemporal Modeling of Fish Trajectory Imputation
Oluwatosin Babasola: A Mathematical Modeling Approach for the Highly Pathogenic Avian Influenza (HPAI) Transmission Dynamics and Cost-Effectiveness Analysis of the Intervention Strategies
Nan Chen: Integrating Data Assimilation with Machine Learning for Advancing Extreme Event Modeling
Kyle Fitch: Improved models for estimating sporadic-E intensity from GNSS radio occultation measurements
Nathan Gaw: Assessing the Calibration and Performance of Attention-based Spatiotemporal Neural Networks for Lightning Prediction
Haiwen Guan: TBA
David Han: AI & Machine Learning for Advanced Nondestructive Evaluation
Tochukwu Ikwunne: Removing Algorithmic Bias and Evaluating Cleft Speech Therapy Game
Sina Khani: Spatial and temporal resolution effects on the structure of extreme events
Victoria Kramb: Training Data Challenges for Nondestructive Inspections (NDI)
Daoji Li: CoxKnockoff: Controlled Feature Selection for the Cox Model Using Knockoffs
Kayode Oshinubi: Spatial variation in climatic factors predicts spatial variation in mosquito abundance in the desert southwest
Abani Patra: TBA
Vishwas Rao: Rare events and their optimization
Alexander Rodríguez: Real-time Disease Outbreak Response by Bridging AI, Data, and Scientific Models
Jianxin Xie: Numerical Differentiation-based Electrophysiology-Aware Adaptive ResNet for Inverse ECG Modeling
Qidong Yang: TBA
Woody Zhu: Optimizing probabilistic conformal prediction with vectorized non-conformity scores
Foundational methods
Talk title: Learning to sample better
Synthesis and outlook
Speakers
Rafal Angryk
Dan Cooley
Elizabeth Gregory
George Karniadakis
Piyush Mehta
Sneha Narra
Kavita Ramanan
Themistoklis Sapsis
Eric Vanden-Eijnden
Jonathan Weare
Rebecca Willett
Tian Zheng
Organizers
Rebecca Willett (co-chair)
George Karniadakis (co-chair)
Richard Davis
Piyush Mehta
Kavita Ramanan
Themistoklis Sapsis
Peter Spaeth
Yao Xie
Tian Zheng
Esteban Real (Google DeepMind): Automatically Discovering Learning Algorithms with Hardware Constraints
Max Nickel (FAIR, Meta AI) – Towards Social Foundations of AI
DSI Summer Programs 2024: Research Symposium
