stochastic optimization

**Stochastic Optimization** is a **class of optimization methods that incorporate randomness in the search process or account for randomness in the objective function** — using probabilistic elements to escape local optima, handle noisy evaluations, and explore large, complex parameter spaces common in semiconductor manufacturing. **Key Stochastic Methods** - **Genetic Algorithms**: Population-based evolution with selection, crossover, and mutation. - **Simulated Annealing**: Random perturbations with temperature-controlled acceptance probability. - **Particle Swarm**: Particles explore the space guided by personal and global best solutions. - **Bayesian Optimization**: Probabilistic surrogate model guides efficient exploration of expensive functions. **Why It Matters** - **Global Optima**: Stochastic methods can escape local optima that trap deterministic gradient methods. - **Noisy Functions**: Naturally handle noisy, stochastic objective functions (yield, process variability). - **No Gradient Needed**: Work with black-box functions where gradients are unavailable. **Stochastic Optimization** is **organized randomness for finding the best** — using controlled randomness to optimize complex, noisy manufacturing processes.