A0311
Title: Quantum enhanced feature subset selection
Authors: Basabi Chakraborty - Iwate Prefectural University (Japan) [presenting]
Abstract: Optimal feature subset selection is an important prerequisite for any pattern classification or machine learning problem. Redundant and irrelevant features degrade performance as well as increase the computational cost of the classifier. An efficient feature evaluation metric and an optimal search process are the two basic requirements for optimal feature subset selection. A lot of feature subset selection algorithms have been developed so far based on statistical and mathematical tools. The recent rapid increase of high dimensional data has created high computational challenges in the area of machine learning and data mining and the need for stable and scalable feature selection algorithms with reduced computational cost is ever increasing. Quantum computing is known to possess enormous processing ability by exploiting inherent parallelism and potentially provides significant speedup compared to classical computing. We would like to describe work on the development of quantum-enhanced optimal feature subset selection algorithms. We have proposed a quantum-based optimization approach to classical feature evaluation metrics and simulated it on classical machines to examine their effectiveness. We have also proposed a novel quantum-inspired metaheuristic based feature selection algorithm and examined its performance with simulation experiments for benchmark data sets.
