Quantum Optimization Approach for Feature Selection in Machine Learning

This is intended to be a technical companion presenting some achievements recently published about the usage of quantum algorithms for the selection of relevant features in a given data set. Based on the paradigm of machine learning, such methods use the concept of mutual information between pairs of observables and between observables and the inferred class, in the special case of a simple classification task. Those probabilistic quantities have been discussed a number of times in several works on information theory. Starting from the paper (Mücke et al., 2023), we provide some further inside about the technical details of their work, with an additional test done on a gate processor using the same binary quadratic approximation model.