Paper-Conference

We identify five factor-graph primitives: a bilinear factor, an exponential link, a Gamma prior, a Gaussian likelihood, and an equality node, and prove that any model composed from them admits closed-form variational message passing

Intelligent agents learn to solve tasks through interactions with their environment. We study autoregressive active inference agents …

We present a probabilistic numerical procedure for optimizing Matérn-class temporal Gaussian processes with respect to the kernel covariance function’s hyperparameters based on Bayesian autoregression.

We present the design of an autoregressive active inference agent in the form of message passing on a factor graph.

We bridge the mathematical and the spike-based perspectives on brain functioning by designing spiking neural networks that simulate Bayesian inference through message passing for Bernoulli messages.

We propose a recursive Bayesian estimation procedure for multivariate autoregressive models with exogenous inputs based on message passing in a factor graph

We propose an active inference agent, consisting of multiple scalar autoregressive model-based agents coupled by virtue of sharing memories, to learn and control a mechanical system with multiple bodies connected by joints.

We describe a Bayesian controller for a cart-pole system, where the entire computational process consists of online Bayesian inference executed by message passing in factor graphs.

For expected free energy minimization, we show that Gaussian approximations that are sensitive to the curvature of the measurement function, such as a second-order Taylor approximation, produce a state-dependent ambiguity term. This induces a preference over states, based on how accurately the state can be inferred from the observation.

We propose a computational procedure for identifying convection in heat transfer dynamics of motion control systems, using a Gaussian Process Latent Force Model.