Quantum computing and quantum-inspired techniques for feature subset selection: a review

Feature subset selection is essential for identifying relevant and non-redundant features, which enhances classification accuracy and simplifies machine learning models. Given the computational difficulties of determining optimal feature subsets, heuristic and metaheuristic algorithms have been widely used. Recently, the rise of quantum computing has led to the exploration of quantum-inspired metaheuristics and quantum-based approaches for this task. Although various studies have explored quantum-inspired and quantum-based approaches for feature subset selection, a comprehensive review that critically examines their significance, limitations, underlying mechanisms, and future directions remains lacking in the literature. This paper addresses this gap by presenting the first in-depth survey of these approaches. We systematically selected and analyzed relevant studies from prominent research databases, providing a detailed evaluation of quantum-inspired metaheuristics and quantum computing paradigms applied to feature subset selection. Our findings indicate that quantum-inspired metaheuristic approaches often deliver superior performance compared to traditional metaheuristic methods for feature subset selection. Nevertheless, their reliance on classical computing limits their ability to fully realize the advantages offered by quantum computing. The quantum-based feature subset selection methods, on the other hand, show considerable promise but are frequently constrained by the current limitations of quantum hardware, making large-scale feature subset selection challenging. Given the rapid evolution of quantum computing, research on both quantum-inspired and quantum-based feature subset selection remains insufficient to draw definitive conclusions. We are optimistic that this review will provide a foundation for future advancements in feature subset selection as quantum computing resources become more accessible.