AIDA: An Active Inference-Based Design Agent for Audio Processing Algorithms

In this paper we present AIDA, which is an active inference-based agent that iteratively designs a personalized audio processing algorithm through situated interactions with human client. The target application of AIDA to propose on-the-spot the most interesting alternative values for tuning parameters hearing aid (HA) algorithm, whenever HA client not satisfied their performance. interprets searching "most alternative" as issue optimal (acoustic) context-aware Bayesian trial design. computational terms, realized Expected Free Energy criterion This type architecture inspired by neuro-economic models on efficient (Bayesian) design in brains and implies comprises generative probabilistic acoustic signals user responses. We novel model sum time-varying auto-regressive filters response based Gaussian Process Classifier. full has been implemented factor graph all tasks (parameter learning, context classification, design, etc.) are variational message passing graph. All verification validation experiments demonstrations freely accessible at our GitHub repository.