Gaussian State Space Model

Gaussian state-space models (GSSMs) are statistical frameworks used to model dynamic systems with latent states evolving over time, observed through noisy measurements. Current research focuses on improving inference methods, particularly for high-dimensional and nonlinear systems, through advancements in particle filtering (including differentiable variants and novel resampling techniques), variational inference, and Markov Chain Monte Carlo (MCMC) methods like Zig-Zag samplers. These improvements aim to enhance the accuracy and efficiency of parameter estimation and state prediction, impacting diverse fields such as climate science, neuroscience, and control engineering through more robust and scalable modeling of complex temporal data. The development of outlier-resistant Kalman filters further expands the applicability of GSSMs to real-world scenarios with noisy or corrupted data.