High-dimensional Optimization

Project Overview

High-dimensional optimization projects focus on optimization problems with high-dimensional input spaces. In such problems, traditional optimization methods often face the curse of dimensionality. Our research solves such high-dimensional optimization problems while improving sampling efficiency and ensuring safety.

Demonstration

MCMC-BO

MCMC-BO is a Bayesian optimization method that leverages Markov Chain Monte Carlo to efficiently sample from an approximated posterior by guiding candidate points toward more promising regions of the search space.

Example

Example for MCMC-BO

HdSafeBO

HDSAFEBO is the first high-dimensional safe Bayesian optimization algorithm that combines optimistic safety identification with isometric embedding to efficiently optimize complex embodied systems while guaranteeing probabilistic safety.

For details, please refer to our project page: HdSafeBO.

HdSafeBO (ours)

HdSafeBO

CONFIG

CONFIG

cEI

cEI

CMAES

CMAES

FocalBO

FocalBO is a hierarchical Bayesian optimization algorithm that trains a focalized sparse GP to zoom-in on promising regions, enabling efficient high-dimensional optimization with large offline or online data.

Example

1-d example for FocalBO

Research Papers

Improving sample efficiency of high dimensional Bayesian optimization with MCMC

Zeji Yi*, Yunyue Wei*, Chu Xin Cheng*, Kaibo He, Yanan Sui

Proceedings of Machine Learning Research (PMLR), 2024

Paper Code

Safe Bayesian Optimization for the Control of High-Dimensional Embodied Systems

Yunyue Wei, Zeji Yi, Hongda Li, Saraswati Soedarmadji, Yanan Sui

Conference on Robot Learning (CoRL), 2024

Paper Project Page Code

Scalable Bayesian Optimization via Focalized Sparse Gaussian Processes

Yunyue Wei, Vincent Zhuang, Saraswati Soedarmadji, Yanan Sui

Neural Information Processing Systems (NeurIPS), 2024

Paper Code
×

Citation