Skip to main content

Conference Program

CompSust DC 2020 will take place online.

Saturday, October 17

11:00am – 12:00pm10:00am – 11:00am8:00am – 9:00am Opening Talk - Introduction to CompSust
Keynote Presentation:
Yoshua Bengio (Mila)
12:00pm – 12:35pm11:00am – 11:35am9:00am – 9:35am
Presentations Group 1
  • Fair Influence Maximization through Welfare Optimization
    Aida Rahmattalabi (University of Southern California)
    ~10 minutes
  • Physically Informed Kernel Learning in Natural Environments
    Victoria Preston (Massachusetts Institute of Technology)
    ~5 minutes
  • Uncertainty-Aware Physics-Informed Neural Networks for Parametrizations in Ocean Modeling
    Björn Lütjens ( Massachusetts Institute of Technology )
    ~5 minutes
  • Towards Global-Scale Species Identification - Scaling Geospatial and Taxonomic Coverage Using Contextual Clues
    Sara M Beery (Caltech)
    ~10 minutes
12:35pm – 12:40pm11:35am – 11:40am9:35am – 9:40am Break
12:40pm – 1:30pm11:40am – 12:30pm9:40am – 10:30am
Presentations Group 2
  • Dual-Mandate Patrols: Multi-Armed Bandits for Green Security
    Lily Xu (Harvard University)
    ~10 minutes
  • Automatic Detection and Compression for Passive Acoustic Monitoring of the African Forest Elephant
    Johan Björck (Cornell)
    ~5 minutes
  • Enhancing Seismic Resilience of Water Pipe Networks
    Taoan Huang (University of Southern California)
    ~5 minutes
  • Multi-objective optimal control for stochastic integrated assessment models
    Angelo Carlino (Politecnico di Milano)
    ~5 minutes
  • Controllable Guarantees for Fair Outcomes via Contrastive Information Estimation
    Umang Gupta (USC Information Sciences Institute)
    ~5 minutes
  • Collapsing Bandits and Their Application to Public Health Interventions
    Aditya S Mate & Jackson Killian (Harvard University)
    ~15 minutes
1:30pm – 2:30pm12:30pm – 1:30pm10:30am – 11:30am Lunch Break & Poster Session
2:30pm – 3:30pm1:30pm – 2:30pm11:30am – 12:30pm Keynote Presentation:
Thomas Dietterich (Oregon State University)
3:30pm – 3:35pm2:30pm – 2:35pm12:30pm – 12:35pm Break
3:35pm – 4:40pm2:35pm – 3:40pm12:35pm – 1:40pm
Presentations Group 3
  • Bayesian Inference for Infectious Disease Outbreaks
    Bryan Wilder (Harvard University)
    ~15 minutes
  • Building Explainable and Interpretable Artificial Intelligence for Solving Decision Problems in Conservation
    Jonathan Ferrer Mestres (CSIRO)
    ~5 minutes
  • Emergency Response Management
    Ayan Mukhopadhyay (Stanford University)
    ~5 minutes
  • Vision for Decisions: Utilizing Real-Time Information from Imagery for Conservation and Public Health
    Elizabeth Bondi (Harvard University)
    ~10 minutes
  • Effects of spatial heterogeneity of leaf density and crown spacing of canopy patches on dry deposition rates
    Theresia Yazbeck (Ohio State University)
    ~5 minutes
  • Mimi.jl – Next Generation Climate Economics Modeling
    Arnav Gautam (Clean Power Research)
    ~10 minutes
  • Synthetic Data Generator for Electric Vehicle Charging Sessions: Modeling and Evaluation Using Real-World Data
    Manu Lahariya (Universiteit Gent)
    ~5 minutes
4:40pm – 4:45pm3:40pm – 3:45pm1:40pm – 1:45pm Break
4:45pm – 5:05pm3:45pm – 4:05pm1:45pm – 2:05pm

Invited Talk: AI + Remote Sensing for Natural Disasters

Ritwik Gupta (Carnegie Mellon Software Engineering Institute)

We can observe the world rapidly and in very high detail using remote sensing assets such as satellites, drones, and balloons. Using artificial intelligence, we can rapidly understand the world as it changes due to natural disasters. I’ll cover what has been done in this area, what is still left to be done, and how people can contribute.
5:05pm – 5:50pm4:05pm – 4:50pm2:05pm – 2:50pm

Tutorial 1: Gaussian Process Regression for Environmental Modeling

Genevieve E Flaspohler (Massachusetts Institute of Technology)

This tutorial will introduce the theory and practice of Gaussian processes regression, with a particular focus on applications in spatiotemporal environmental modeling. Topics covered will include correlation and covariance, multivariate Gaussian models, Gaussian conditioning and inference, Gaussian processes, kernel and mean functions, and maximum likelihood estimation. The tutorial will include a hands-on, coding portion in which participants will be able to fit a GP model to an environmental dataset in the Western US and examine the properties of the resulting model. Prior exposure to introductory probability/statistics --- especially Gaussian distributions and correlation/covariance --- is useful but not necessary for the tutorial
5:50pm – 6:00pm4:50pm – 5:00pm2:50pm – 3:00pm Break
6:00pm – 6:55pm5:00pm – 5:55pm3:00pm – 3:55pm Collaborathon: Team Generation
6:55pm – 8:00pm5:55pm – 7:00pm3:55pm – 5:00pm Prep for Evening Networking Activity & Dinner
8:00pm7:00pm5:00pm Dinner Party/Networking Activity

Sunday, October 18

11:00am – 12:00pm10:00am – 11:00am8:00am – 9:00am Keynote Presentation: AI for Public Health and Conservation: Learning and Planning in the Data-to-Deployment Pipeline
Milind Tambe (Harvard University)
12:00pm – 12:35pm11:00am – 11:35am9:00am – 9:35am
Presentations Group 4
  • Electricity Systems: Tackling Climate Change with Machine Learning
    Priya Donti (Carnegie Mellon University)
    ~15 minutes
  • Integrating Machine Learning and Numerical Weather Prediction
    Genevieve E Flaspohler (Massachusetts Institute of Technology)
    ~5 minutes
  • Sustainable Claim Matching for Fact Checkers
    Benjamin I Rocklin (Stanford)
    ~5 minutes
  • Message passing neural network for predicting CO2 adsorption in metal-organic frameworks (MOFs)
    Ali Raza (Oregon State University)
    ~5 minutes
12:35pm – 12:40pm11:35am – 11:40am9:35am – 9:40am Break
12:40pm – 1:35pm11:40am – 12:35pm9:40am – 10:35am

Tutorial 2: Building Models for Static Sensors: the good, the bad, and the ugly

Sara Beery (Caltech)

Ecological data is frequently collected from static sensors, like camera traps, acoustic receivers like AudioMoth, or static sonar used to monitor species underwater. This data presents challenges that are not well addressed by existing machine learning methods, including a large amount of "empty" data, a small number of examples for most species, strong and often spurious correlations across data collected from one sensor installation, and highly variable signal quality. In this tutorial, we will discuss some of the ways to adapt existing methods to handle these challenges and get hands-on with a real-world dataset to determine how to best structure the data for training and evaluation of ML methods.
1:35pm – 2:30pm12:35pm – 1:30pm10:35am – 11:30am Lunch Break & Poster Session
2:30pm – 3:05pm1:30pm – 2:05pm11:30am – 12:05pm
Presentations Group 5
  • The Impact of Risk Transfer Mechanisms on Smallholder Farmer Climate Adaptation
    Nicolas Choquette-Levy (Princeton University)
    ~10 minutes
  • Neural Arbors are Pareto Optimal
    Arjun Chandrasekhar (University of Pittsburgh)
    ~5 minutes
  • Optimization of Ride-Hailing Electrification Considering Emissions Costs
    Matthew B Bruchon (Carnegie Mellon University)
    ~5 minutes
  • Reducing the world with osm-tag-stats
    Aruna Sankaranarayanan (Massachusetts Institute of Technology)
    ~10 minutes
3:05pm – 3:10pm2:05pm – 2:10pm12:05pm – 12:10pm Break
3:10pm – 4:10pm2:10pm – 3:10pm12:10pm – 1:10pm
Presentations Group 6
  • Adaptive-Halting Policy Networks for Early Classification
    Thomas Hartvigsen (Worcester Polytechnic Institute)
    ~10 minutes
  • Differentiable Optimal Adversaries for Learning Fair Representations
    Aaron M. Ferber (University of Southern California)
    ~5 minutes
  • AI for Food Security and Bandit Data-driven Optimization
    Zheyuan Ryan Shi (Carnegie Mellon University)
    ~10 minutes
  • Evaluating the Fairness of Bike Sharing Programs with Geospatial Analysis
    Katelyn Morrison (University of Pittsburgh)
    ~5 minutes
  • Machine Learning for Material Defect Analysis with Minimal Human Input
    Md Nasim (Purdue University)
    ~5 minutes
  • deepGEFF: Forecasting Wildfire Danger and Spread with Deep Learning
    Anurag Saha Roy & Roshni Biswas (Wikilimo)
    ~10 minutes
  • Statistical and machine learning methods for evaluating emissions reduction policies on air quality under changing meteorological conditions
    Minghao Qiu (Massachusetts Institute of Technology)
    ~10 minutes
4:10pm – 4:20pm3:10pm – 3:20pm1:10pm – 1:20pm Break
4:20pm – 5:05pm3:20pm – 4:05pm1:20pm – 2:05pm

Tutorial 3: Smart Emergency Response : Forecasting, Resource Allocation, and Deployment

Ayan Mukhopadhyay (Stanford University)

Emergency response to incidents such as accidents, medical calls, and fires is one of the most pressing problems faced by communities across the globe. In the last fifty years, researchers have developed statistical, analytical, and algorithmic approaches for designing emergency response management (ERM) systems. ERM comprises of three fundamental problems -- spatial-temporal incident prediction, resource allocation, and dispatch. In this tutorial, we will seek to understand how to tackle these problems. First, we will discuss how data-driven models can be learned to forecast spatial temporal incidents. We will also explore how to account for robustness in such approaches. Then, given such models, we will look at how agents can be deployed in anticipation of incidents. Finally, we will explore algorithmic approaches to dispatching resources.
5:05pm – 5:10pm4:05pm – 4:10pm2:05pm – 2:10pm Break
5:10pm – 6:00pm4:10pm – 5:00pm2:10pm – 3:00pm Collaborathon Team Presentations
6:00pm – 6:50pm5:00pm – 5:50pm3:00pm – 3:50pm

Panel Discussion: Computational Sustainability: Research, Career, and Future

Panelists: Yoshua Bengio, Carla Gomes, Douglas Fisher, Rebecca Hutchinson, Thomas Dietterich
6:50pm – 7:00pm5:50pm – 6:00pm3:50pm – 4:00pm Closing Remarks

Keynote speakers

Yoshua Bengio

Yoshua Bengio (Mila)

Sat Oct 17
11am EDT / 10am CDT / 8am PDT

Thomas Dietterich

Thomas Dietterich (Oregon State University)

Machine Learning and Computational Sustainability: Lessons Learned and Future Challenges

Sat Oct 17
2:30pm EDT / 1:30pm CDT / 11:30am PDT

Abstract: When Carla Gomes and I first launched the CompSust effort, I often described ecological sustainability applications in terms of six steps: sensor placement (data collection), data interpretation, data integration, model fitting, policy optimization, and policy execution. In this talk, I will revisit this pipeline and discuss the lessons learned from our research at Oregon State and the challenges for the future.

Bio: Dr. Dietterich is Professor Emeritus in the School of Electrical Engineering and Computer Science at Oregon State University. Dietterich has served as co-PI and Assistant Director of the Institute for Computational Sustainability, which has led the two NSF Expedition grants in Computational Sustainability. He has supervised or co-supervised several postdocs and PhD students in this area including Rebecca Hutchinson (OSU), Dan Sheldon (UMass), Mark Crowley (Waterloo), Yann Dujardin (INRA Auzeville, France), Jonathan Ferrer (CSIRO, Brisbane), Liping Liu (Tufts), Sean McGregor (X-PRIZE), Majid Alkaee Taleghan (eSentire), Kim Hall (OSU), Tadesse Zemicheal (NVIDIA), Amelia Snyder (World Resources Institute).

Milind Tambe

Milind Tambe (Harvard University)

AI for Public Health and Conservation: Learning and Planning in the Data-to-Deployment Pipeline

Sun Oct 18
11am EDT / 10am CDT / 8am PDT

Abstract: With the maturing of AI and multiagent systems research, we have a tremendous opportunity to direct these advances towards addressing complex societal problems. We focus on the problems of public health and wildlife conservation, and present research advances in multiagent systems to address one key cross-cutting challenge: how to effectively deploy our limited intervention resources in these problem domains. We present our deployments from around the world as well as lessons learned that we hope are of use to researchers who are interested in AI for Social Impact. Achieving social impact in these domains often requires methodological advances; we will highlight key research advances in topics such as computational game theory, multi-armed bandits and influence maximization in social networks for addressing challenges in public health and conservation. In pushing this research agenda, we believe AI can indeed play an important role in fighting social injustice and improving society.

Bio: Milind Tambe is Gordon McKay Professor of Computer Science and Director of Center for Research in Computation and Society at Harvard University; concurrently, he is also Director "AI for Social Good" at Google Research India. He is a recipient of the IJCAI John McCarthy Award, ACM/SIGAI Autonomous Agents Research Award from AAMAS, AAAI Robert S Engelmore Memorial Lecture award, INFORMS Wagner prize, Rist Prize of the Military Operations Research Society, Christopher Columbus Fellowship Foundation Homeland security award, AAMAS influential paper award, best paper awards at conferences including AAMAS, IJCAI, IVA. He has also received meritorious commendations and letters of appreciation from the US Coast Guard, Los Angeles Airport, and the US Federal Air Marshals Service. Prof. Tambe is a fellow of AAAI and ACM.

Steering Committee