Speaker
Description
Lévy walk is a non-Markovian anomalous diffusion model that has been widely used to describe random walks in diverse fields such as biology, quantum physics, and ecology. One of the controversial issues in these fields is that whether the given diffusion dynamics can be successfully explained by a Lévy walk or not, which becomes a nontrivial issue if diffusion data is short and noisy. In this work, we propose a Bayesian inference method for model classification and parameter estimation of time trace data of noisy Lévy walks. We propose how to calculate the likelihood function of a Lévy walk, which is a non-Markovian process, via hidden Markov models. With this likelihood function, we carry out the Bayesian model comparison and parameter estimation on the simulated trajectories at various conditions. We demonstrate that the likelihood function is successfully calculated via hidden Markov model, and the Bayesian inference works well for the trajectories with moderate signal-to-noise levels.
