Course Schedule
Live Schedule on Google Sheets
The schedule below is tentative. We may spend more or less time on a topic depending on student interest.
Weekly Topics and Readings
| Date | Topic | Required Paper 1 | Required Paper 2 | Optional / Further Reading | Deliverables |
|---|---|---|---|---|---|
| 1/6/2026 | Intro: importance of constraints + classical approaches | Adopt Constraints Over Penalties in Deep Learning (Ramirez et al., 2025) [arXiv] | Optimization Learning (Van Hentenryck, 2025) [arXiv] | Fioretto & Van Hentenryck (2020) [arXiv] | |
| 1/13/2026 | No class; read ahead | — | — | Students select ≥2 papers from Weeks 3–9 for short presentation proposals | Presentation proposal: 5-10 papers and initial (AI generated or better) presentation |
| 1/20/2026 | Differentiable projections & feasibility layers | HardNet: Hard-Constrained Neural Networks (Min & Azizan, 2025) [arXiv] | OptNet: Differentiable Optimization as a Layer (Amos & Kolter, 2017) [arXiv] | CVXPYLayers [arXiv]; DC3; Enforcing Hard Linear Constraints with Decision Rules (Constante-Flores et al., 2025) [arXiv] | |
| 1/27/2026 | Reinforcement learning with constraints | Trust Region Policy Optimization (Schulman et al., 2015) [arXiv] | Constrained Policy Optimization (Achiam et al., 2017) [arXiv] | Projection-Based CPO (PCPO) [arXiv]; FOCOPS [arXiv]; Distributionally Robust Constrained RL (Zhang et al., 2024) [arXiv] | |
| 2/3/2026 | Safe RL: Lyapunov & reachability | Lyapunov-based Safe RL [arXiv] | Robust Constrained Reinforcement Learning (Zheng et al, 2024) [arXiv]) | Control barrier functions; MPC + learning; robust RL formulations 1 | Project proposal |
| 2/10/2026 | LLM alignment | Training Language Models with Human Feedback (RLHF) (Ouyang et al., 2022) [arXiv] | Constitutional AI (Bai et al., 2022) [arXiv] | Safety Representation Ranking (SRR) [OpenReview]; critiques of reward modeling | |
| 2/17/2026 | Post-training interventions & steering | Constrained Decoding (Hokamp & Liu, 2017) [arXiv] | Persona Vectors (Chen et al., 2025) [arXiv] | AlphaEdit [arXiv]; Guiding LLMs the Right Way [arXiv]; Surjectivity of Neural Networks [arXiv] | |
| 2/24/2026 | Formal verification | Vericoding / Verified Code Generation [survey] | Hilbert: Informal + Formal Proofs (Varambally et al., 2025) [arXiv] | Lean-guided LLMs; proof-carrying code; formal methods for agents | Project midterm report |
| 3/3/2026 | Declarative programming & tool use | OptiMUS-0.3 (AhmadiTeshnizi et al., 2025) [arXiv] | OptiMind: Teaching LLMs Optimization (Chen et al., 2025) [MSR] | OptiChat [arXiv]; LLMs for supply chain decisions [arXiv]; Democratizing Optimization with GenAI | |
| 3/10/2026 | Synthesis & future directions | On Surjectivity of Neural Networks (Jiang & Haghtalab, 2025) [arXiv] | GDPval: Economically Grounded Evaluation (Patwardhan et al., 2025) [arXiv] | SWE-bench [arXiv]; agent configuration constraints [arXiv] | |
| 3/18/2026 | (No class; exam period) | — | — | — | Project final report |