Active recursive Bayesian inference using Rényi information measures

• An active recursive inference method based on Rényi information measures. Posterior probability changes can enhance both query optimization and stopping criterion in the RBI process. Using entropy α -divergence unifies multi-objective problem. The proposed unified framework enhances speed accuracy, presence of an adversarial prior. Recursive Bayesian (RBI) provides optimal latent variable estimates real-time settings with streaming noisy observations. Active attempts to effectively select queries that lead more informative observations reduce uncertainty until process is stopped at a certain confidence level. However, mismatch between querying objective creates conundrum improving performance one expense deteriorating other. Moreover, conventional methods stagger misleading prior information. Inspired by theoretic approaches, we propose where are jointly selected through formulation enables us accuracy We theoretically demonstrate encourages exploration Furthermore, motivate our proving geometrical representation for decision making simplex. provide empirical experimental studies two applications including restaurant recommendation brain-computer interface (BCI) typing systems outperforms comparable approaches accuracy.