Bayesian Feature Selection for Hearing Aid Personalization

We formulate hearing aid personalization as a linear regression. Since sample sizes may be low and the number of features may be high we resort to a Bayesian approach for sparse linear regression that can deal with many features, in order to find efficient representations for on-line usage. We compare to a heuristic feature selection approach that we optimized for speed. Results on synthetic data with irrelevant and redundant features indicate that Bayesian backfitting has labelling accuracy comparable to the heuristic approach (for moderate sample sizes), but takes much larger training times. We then determine features for hearing aid personalization by applying the method to hearing aid preference data.