Sanjeev Arora, Princeton University
Nina Balcan, Carnegie-Mellon University
Sham Kakade, University of Washington
Sanjoy Dasgupta, University of California, San Diego
The second Simons Symposium on New Directions in Theoretical Machine Learning will focus on the following questions:
- Does lack of fundamental understanding hold back progress toward the goal of AI and general purpose learning agents?
- What kinds of new fundamental models and understanding are needed?
- Is current theory on the right path?
Agenda & Slides
MONDAY, SEPTEMBER 5
Chris Manning | Compositionality (in LMs and Generalization) View Slides (PDF) Nati Srebro | Benign Overfitting and Why it Matters View Slides (PDF) Seb Bubeck | Unveiling Transformers with LEGO View Slides (PDF) Sasha Rush | Quantifying LM Behaviors: Prompting and Discourse View Slides (PDF) Tengyu Ma | Three Facets of Understanding Pre-Training: Self-Supervised Loss, Inductive bias and Implicit Bias View Slides (PDF)
TUESDAY, SEPTEMBER 6
Yejin Choi | The Neuro-Symbolic Continuum Between Language, Knowledge and Reasoning View Slides (PDF) Stefanie Jegelka | Representation and Learning in Graph Neural Networks - An Overview of Results View Slides (PDF) Surbhi Goel | Sparse Feature Emergence in Deep Learning View Slides (PDF) Jason Lee | Beyond NTK via Feature Learning with SGD View Slides (PDF) Sanjoy Dasgupta | Memory Games View Slides (PDF)
WEDNESDAY, SEPTEMBER 7
THURSDAY, SEPTEMBER 8
Jitendra Malik | Adaptive Control via Deep RL, with Applications to Robotics View Slides (PDF) Sham Kakade | Why Reward is Not Enough for RL View Slides (PDF) Dylan Foster | The Statistical Complexity of Interactive Decision Making View Slides (PDF) Jamie Morgenstern | Shifts in Distributions and Preferences in Response to Learning View Slides (PDF) Samory Kpotufe | Tracking Significant Changes in Bandits View Slides (PDF)
FRIDAY, SEPTEMBER 9
Roger Grosse | How to Approximate Neural Net Function Space Distance and Why View Slides (PDF) Sanjeev Arora | Deep Learning Becoming Even More of a Black Box: What Can Theory Do? View Slides (PDF) Surya Ganguli | Beyond Neural Scaling Laws: Beating Power Law Scaling Through View Slides (PDF) Ellen Vittercik | Theoretical Foundations of Machine Learning for Cutting Plane Selection View Slides (PDF) Simon Du | Offline Reinforcement Learning View Slides (PDF)