Paper: Solving Mixed-Integer Optimization with Constraint-Aware Diffusion
nandofioretto · x · 2026-08-30
This paper proposes Constrained Graph Diffusion (CGD), a framework using graph-based diffusion models to generate discrete decisions in Mixed-Integer Programming (MIP). It integrates a training-free feasibility projection operator into the reverse diffusion process to maintain validity. Once discrete decisions are set, the remaining continuous subproblem is solved numerically. Evaluations on optimal transmission switching and portfolio optimization show substantial improvements in feasibility and quality, achieving speedups of up to 425x over state-of-the-art solvers.
More from Research
- Schmidhuber: I Published the First Concrete Recursive Self-Improvement Algorithms Back in 1987 — SchmidhuberAI · 2026-09-23
- Quanta Explains How Pricing Algorithms Can Drive Up Prices Without Collusion — burny_tech · 2026-09-23
- CodeMidas: Turning Raw Source Code into Executable RL Environments for Coding Agents — burny_tech · 2026-09-23
- Researchers pine for pre-2000s methodological papers with no mathiness or defensive fluff — PMinervini · 2026-09-23
- Block-triangular joint drifting enables one-step generative surrogate models for stochastic trajectories — chaumian · 2026-09-23
- Raw LLM probabilities aren't enough for decisions — calibration matters, researchers argue — PMinervini · 2026-09-23