The two-day program included keynote talks, invited speakers, and discussions. The conference also provided ample opportunities to meet, convene, and converse with colleagues and attendees.
Tuesday, May 21, 2024
8:30 – 8:40 Welcome Message, Alexandre Bayen, Director, CITRIS and the Banatao Institute, University of California, Berkeley [video, slides]
8:40 – 9:00 Workshop Introduction and Goals, Conference Organizers
9:00 – 9:40 Neural Certificates
Thomas Henzinger, Institute of Science and Technology Austria (Invited Plenary) [video, slides]
9:40 – 10:00 Break
Session 1: Foundation Models and Neuro-Symbolic Systems Analysis
Moderator: Alvaro Velasquez, Defense Advanced Research Projects Agency
10:00 – 10:20 Emergence of Symbolic Computation in Stochastic Realizations of Large Scale Generative Models: Glimpses of Solomonoff Through Holes on the Markov Blanket, Stefano Soatto, University of California, Los Angeles and Amazon Web Services AI [video]
10:20 – 10:40 LLMs Can't Plan, But Can Help Planning in LLM-Modulo Frameworks, Subbarao Kambhampati, Arizona State University [video, slides]
10:40 – 11:00 Confidence Metrics using Bayesian Hypothesis Testing and Semantic Perturbations of Symbolic Universal Scene Descriptions, Sumit Jha, Florida International University [video, slides]
11:00 – 11:20 Heat Death of Generative Models in Closed-Loop Learning, Paulo Tabuada, University of California, Los Angeles [video, slides]
11:20 – 11:40 Creative Generative Models for Scientific Discovery of Cyberphysical Systerms Design, Susmit Jha, SRI International [video, slides]
11:40 – 12:40 Lunch
12:40 – 1:20 Interpretable AI via Neuro-Symbolic Information Pursuit
René Vidal, University of Pennsylvania (Invited Plenary) [video, slides]
Session 2: Reconciling Deep Learning with Physics-Based Models
Moderator: George Pappas, University of Pennsylvania
1:30 – 1:50 Bridging Numerical Methods and Deep Learning with Physics-Constrained Differentiable Solvers, Aditi Krishnapriyan, University of California, Berkeley [video, slides]
1:50 – 2:10 Theorem Proving using Language Models and Lean, Anima Anandkumar, California Institute of Technology [video, slides]
2:10 – 2:30 Neural Network Approximations of Solutions of Nonlinear Hyperbolic Conservation Laws, Alexandre Bayen, University of California, Berkeley [video, slides]
2:30 – 2:50 Towards Multi-Sensory World Models, Antonio Loquercio, University of Pennsylvania [video, slides]
2:50 – 3:10 Design of Autonomous Neurosymbolic Agents using Evolving Behavior Trees, Xenofon Koutsoukos, Vanderbilt University [video, slides]
3:10 – 3:30 Assured NN-based Perception using Physics-based Generative Models, Yasser Shoukry, University of California, Irvine [video, slides]
3:30 – 4:00 Break
Session 3: Neuro-Symbolic Design Tools
Moderator: Sanjit Seshia, University of California, Berkeley
4:00 – 4:20 Safety Analysis of AI-enabled Cyber-Physical Systems: A Formal Approach, Pavithra Prabhakar, Kansas State University [video, slides]
4:20 – 4:40 Synthesis and Verification of Neural Feedback Controllers for Temporal Logic Tasks, Georgios Fainekos, Toyota Motor North America, Research & Development [video, slides]
4:40 – 5:00 Hardware-Software Co-Design for Edge Computing, Helen Li, Duke University [video, slides]
5:00 – 5:20 Unlocking the "Blackbox" of Data-Driven Optimization for Energy Systems Planning, Saurabh Amin, Massachusetts Institute of Technology [video, slides]
5:20 – 6:00 Design Automation and Neuro-Symbolic AI Systems
Sanjit Seshia, University of California, Berkeley (Invited Plenary) [video, slides]
6:00 – 7:30 Networking Reception
Wednesday, May 22, 2024
8:30 – 8:50 Reflections on Day 1 [video]
9:00 – 9:40 How the Brain Represents the World: Insights from Reading the Brain’s Code for Face
Doris Tsao, University of California, Berkeley (Invited Plenary) [video, slides]
9:40 – 10:00 Break
Session 4: Analysis Methods
Moderator: Rajeev Alur, University of Pennsylvania
10:00 – 10:20 Formal Verification for Neuro-Symbolic Systems: Verifying Safety and Liveness in Neuro-Symbolic Behavior Trees (NSBTs), Taylor Johnson, Vanderbilt University [video, slides]
10:20 – 10:40 Assured Neuro-Symbolic Learning-based Control Subject to Timed Temporal Logic Constraints, Kyriakos Vamvoudakis, Georgia Institute of Technology [video, slides]
10:40 – 11:00 Wavelet-inspired Neural Operators as Foundations for Designing Neuro-Symbolic Systems: Discovering the Governing Equations, Paul Bogdan, University of Southern California [video, slides]
11:00 – 11:20 Neuromorphic Computing with Symbols, Rajit Manohar, Yale University [video, slides]
11:20 – 11:40 Towards a Neuro-Symbolic Analysis of Human-Machine Teams, Michael Warren, HRL Laboratories [video, slides]
11:40 – 12:00 Neuro-Symbolic Methods for Autonomous Agents, Alberto Speranzon, Lockheed Martin Advanced Technology Labs [video, slides]
12:00 – 12:40 Lunch
12:40 – 1:20 Bridging the Gap between Learning and Programming
Armando Solar-Lezama, Massachusetts Institute of Technology (Invited Plenary) [video, slides]
Session 5: Integrating Design and Analysis for Assurance
Moderator: Larry Rohrbough, University of California, Berkeley
1:30 – 1:50 Neurosymbolic Methods for Estimating Common Ground and Semantic Structures for Human Machine Teams, Eric Davis, Defense Advanced Research Projects Agency [video, slides]
1:50 – 2:10 Regular Reinforcement Learning, Ashutosh Trivedi, University of Colorado Boulder [video, slides]
2:10 – 2:30 Towards Neuro-Symbolic Video Understanding, Sandeep Chinchali, University of Texas at Austin [video, slides]
2:30 – 2:50 Neuro-Symbolic Policy Learning and Neuro-Symbolic Verification, Jyotirmoy Deshmukh, University of Southern California [video, slides]
2:50 – 3:10 Neuro-Causal Models, Pradeep Ravikumar, Carnegie Mellon University [video, slides]
Session 6: Design Tools
Moderator: Shankar Sastry, University of California, Berkeley
3:10 – 3:30 Towards Neuro-Symbolic Safety Certification of Human-Centered Systems, Jaime Fernández Fisac, Princeton University [video, slides]
3:30 – 3:50 Auto-Encoding Bayesian Inverse Games, David Fridovich-Keil, University of Texas at Austin [video, slides]
3:50 – 4:10 Exploiting Symbolic Structures for Optimal Control and Decision Making, Arun Ramamurthy, Siemens [video, slides]
4:10 – 4:30 Neurosymbolic Reasoning Challenges in Safe Vehicle Operations in Unknown Environments, Sanjai Narain, Peraton Labs [video, slides]
4:30 – 4:50 Integration of AI/ML with MBSE for Trusted Autonomy, John Baras, University of Maryland [video, slides]
Session 7: NeuS 2025 Planning
5:00 – 6:00 Open Problems, Future Directions [video]
6:00 Closing Remarks and Wine/Cheese
All times are Pacific Daylight Time.