Learning Where to Park by Active Inference

In order to assess the feasibility of active inference as a framework for synthetic agents in a real-world setting, we developed a ground-based robot that needs to learn to navigate to an undisclosed parking location. The robot can only learn where to park through situated interactions with a human observer who is aware of the target location.

Read more

ForneyLab: a Julia Toolbox for Factor Graph-Based Probabilistic Programming

ForneyLab is a novel Julia package that allows the user to specify a probabilistic model as an FFG and pose inference problems on this FFG. ForneyLab is especially potent when applied to time-series data, where it attains comparable performance to Stan and Edward in significantly less computation time.

Read more

Bayesian Pure-tone Audiometry

We develop a fully probabilistic approach to pure-tone audiometry. By assuming a Gaussian process based response model for test tones, the hearing threshold estimation problem becomes one of Bayesian inference. This allows the use of information-theoretic criteria to select optimal test tones.

Read more