Automated Algorithm Configuration

  
Home  |  Abstract  |  People  |  Software  |  Licensing  |  Achieved

Abstract

Automated algorithm configuration (AAC), which is also known as hyper-parameter optimization, aims to seek high-performance configurations for a given algorithm/system, and is a key component in automated algorithm design and automated machine learning (AutoML).

In this project, we develop a powerful AAC technique based on neural network enhancement, and successfully apply it to significantly pushing forward the state of the art in solving the well-known NP-hard combinatorial optimization problem of minimum vertex cover (MinVC). MinVC is an influential problem in graph theory with extensive real-world applications.

Based on the idea of automated algorithm design, we develop and present a powerful MinVC solver, dubbed MetaVC, which is a highly parametric local search framework for MinVC.

MetaVC incorporates various techniques that are automatically customized and combined, using effective automated algorithm configuration, for effectively solving MinVC instances.

Papers

Software

Licensing

MetaVC is under MIT License.

Please send any questions, concerns or comments to Chuan Luo.

Achieved

Return to homepage.