Description: Derivative-free optimization methods are suitable for sophisticated optimization problems, while are hard to scale to high dimensionality (e.g., larger than 1,000). Previously, the random embedding technique has been shown successful for solving high-dimensional problems with low effective dimensions. But for high-dimensional problems without low effective dimensions, there exists embedding gap when applying random embedding. Therefore, sequential random embeddings (SRE) technique is proposed for high-dimensional problems with low optimal epsilon-effective dimensions. SRE can reduce the embedding gap while still running optimization algorithms in the low-dimensional spaces. Theoretical and empirical results show the effectiveness of SRE for the high-dimensional non-convex optimization problems with low optimal epsilon-effective dimensions.
SRE-framework is an abstract algorithm framework. A specific optimization algorithm is an essential part for this framework and should be implemented. SRE-framework can be used to solve high-dimensional non-convex optimization problems with low optimal epsilon-effective dimensions.
Reference: Hong Qian, Yi-Qi Hu, and Yang Yu.
Derivative-free optimization of high-dimensional non-convex functions by sequential random embeddings. In Proceedings of the 25th International Joint Conference on Artificial Intelligence (IJCAI’16), New York, NY, 2016.
ATTN: This package is free for academic usage. You can run it at your own risk. For other purposes, please contact
Prof. Zhi-Hua Zhou.
Requirement: The package was developed with Java.
ATTN2: This package was developed by Mr. Yi-Qi Hu and Dr. Yang Yu. For any problem concerning the code, please feel free to contact
Mr. Hu or
Dr. Yu.
Download:
code (24Kb)