Theory and Control
summary
Four papers: a neural method that learns Signal Temporal Logic formulas by training a recurrent net whose weights and one‑hot choice blocks encode both structure and parameters, scales to tens of thousands of traces, yields equal or better misclassification rates but much longer training; an algorithm that under‑approximates the Minkowski difference of constrained zonotopes via a polynomial‑size linear program, implemented in MATLAB and reported as scalable to high‑dimensional nonlinear backward reachability; an integer linear program with iterative path‑constraint generation that finds the minimum WCRT for DAG tasks in minutes and typically needs ≈13% of paths; and a new critical‑chain DP response‑time bound for limited‑preemptive sporadic DAGs that outperforms prior bounds on benchmarks.
AI-written summary · about
chapters (15)
- 0:00Start
- 1:05Differential Inference of Signal Temporal Logic
- 2:31Logical Inference Method
- 7:45Choice Blocks
- 24:47Problem Formulation
- 25:17Constant Result
- 29:36Experimental Results
- 38:31Summary of the Irp Model
- 39:10Challenge Problem
- 39:38Inter Iterative Solving Method
- 41:10Evaluation Results
- 42:05Conclusion
- 52:28Lower Product Interference
- 55:29Dynamic Programming Algorithm
- 56:41Fixed Priority Scheduling Policy
related talks
01h 07mBack to the Future: A Fresh Look at Linear Temporal Logic
01h 02mLogic, Optimization, and Constraint Programming: A Fruitful Collaboration
19mSMT Theory Arbitrage: Approximating Unbounded Constraints using Bounded Theories
19mPanel: Compilation Frameworks
17mCode World Model: Building World Models for Computation
37mJourney with Eric Allender: From Turing Machines to Circuits, and Back
29mA Temporal Differential Dynamic Logic Formal Embedding
40mReasoning Under Uncertainty in SMT Solving, Research, and Life