Optimization for One, Feasibility for Most: Decoupling Optimization from Feasibility in Chance-Constrained Programs

Key points

• Chance-constrained stochastic programs (CCSPs) require solutions that satisfy uncertain constraints with high probability.1
• Standard approaches either embed scenarios directly into optimization by considering sample average approximation (SAA)-based reformulations, growing prohibitively expensive and often failing to find feasible solutions; or employ conservative reformulations that guarantee feasibility but sacrifice objective quality by design.1
• First, high-quality candidate solutions are generated by solving many inexpensive single-scenario subproblems, avoiding sampling directly in the constrained high-dimensional solution space.1
• We also introduce a variant, ACCESS($δ$), which provides certified feasibility guarantees with probability $1-δ$.1

Sources (1)

• [1]Optimization for One, Feasibility for Most: Decoupling Optimization from Feasibility in Chance-Constrained Programs
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 13, 06:20 PM

  We propose ACCESS (Adaptive Chance-Constrained Exploration via Structured Sampling), a structured sampling framework that decouples optimization from feasibility evaluation.

  Chance-constrained stochastic programs (CCSPs) require solutions that satisfy uncertain constraints with high probability.

Extractive summary: sentences quoted from the sources.