Tractable inference and observation likelihood evaluation in latent structure influence models
Published in IEEE Transactions on Signal Processing, 2020
This work introduces efficient algorithms for inference and observation likelihood evaluation in Latent Structure Influence Models (LSIMs). The paper addresses fundamental computational challenges in coupled hidden Markov models by developing tractable approximation methods that maintain high accuracy while significantly reducing computational complexity.
The proposed framework enables practical implementation of LSIMs for real-time signal processing applications, particularly in scenarios involving multi-channel time series data. The methods presented have broad implications for biomedical signal processing, financial modeling, and other domains requiring analysis of coupled temporal processes.
Recommended citation: Karimi, S., & Shamsollahi, M. B. (2020). "Tractable inference and observation likelihood evaluation in latent structure influence models." IEEE Transactions on Signal Processing, 68, 5736-5745.
Download Paper