Eighth Annual Industry Day
2022-11-16
Berkeley, United States
who it's for
machine learning researchers, applied machine learning engineers, reinforcement learning researchers, data engineers for large-scale pipelines, security and cryptography engineers, computational transportation engineers, fraud detection engineers
topics
zero-knowledge SNARKs as a service; MPC-in-the-exponent protocols; Groth-16 prover optimizations; large road graph partitioning; fast Laplacian solvers for routing; compact distance oracles; EV routing with battery and charger state; model-agnostic RL explainers with decomposed reward forecasts; adversarial example generation for transaction data; adversarial retraining and hard-example defenses; factored policy gradient with influence networks; document OCR with YOLO and CNN+k-means
| title | author | length |
|---|---|---|
| Introduction and Research Fellows Lightning Talks | Shafi Goldwasser, Omid Farah, Kalani Pendland, Carson Young, William Homnicki, Amy Ambrose | 49m |
Five-minute snapshots present concrete claims and tools: queueing work gives implementable rules (how many hardware priority buckets and handling unknown… | ||
| zkSaaS: Zero-Knowledge SNARKs as a Service | Sanjam Garg | 18m |
Builds a distributed protocol that secret-shares a witness across one large host and many commodity workers, runs an MPC in the exponent to avoid blowing up… | ||
| Computational Mobility: Learning, Prediction, and Algorithms | Sreenivas Gollapudi | 15m |
Describes production-ready methods for large road graphs: partitioning via random walks plus max flow yields balanced separators that cut query latency and are… | ||
| CrystalBox: Efficient Model-Agnostic Explanations for Deep RL Controllers | Nina Narodytska | 17m |
CrystalBox trains a neural black-box model-agnostic explainer on replayed system traces (eg Puffer) by labeling state+action pairs with horizon-wise… | ||
| Generating Adversarial Examples & Defense Methods For Online Fraud Detection | Nitin Sharma | 19m |
Claims that realistic adversarial examples for PayPal-style transaction fraud can be generated by combining ensemble adversarial perturbations (FGSM, L-BFGS… | ||
| Sequential Decision Making Problems in Multi-Agent Environments | Sumitra Ganesh | 15m |
Group presents a method called factored policy gradient that uses an influence network (bipartite graph linking action dimensions to reward components) and a… | ||
| Unique Challenges and Opportunities in Working with Residential Real Estate Data | Takamitsu Tanaka | 16m |
Residential real estate data is messy and nonstandard across counties, so the team builds pipelines that scan public records, run YOLO object detection to pick… | ||
| Highlights from Current Programs part 2: | Siddhartha Banerjee, Christian Borgs, Avishay Tal | 24m |
Three short programs report concrete tools and open claims: a graph limit (graphon) model that generates sparse networks from two-feature link functions and… | ||
| nothing matches | ||