New Bayesian optimization method minimizes worst-case functional errors

Researchers have introduced a new framework called min-max Functional Bayesian Optimization (MM-FBO) to address challenges in optimizing functions with functional responses, which are common in scientific and engineering fields. Unlike existing methods that focus on average performance, MM-FBO directly minimizes the maximum error across the entire functional domain. The approach represents functional responses using functional principal component analysis and employs Gaussian process surrogates to manage uncertainty, balancing exploitation of worst-case errors with exploration. Experiments on synthetic and physics-based benchmarks demonstrate MM-FBO's superior performance compared to current baselines. AI

影响 Introduces a novel optimization technique for complex functional responses, potentially improving efficiency in scientific and engineering simulations.

排序理由 This is a research paper detailing a new optimization framework.

在 arXiv stat.ML 阅读 →

• arXiv
• Gaussian process
• Pouya Ahadi

• 论文
• 其他

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →