HGF Toolbox v2.1

Version 2.1 of the HGF Toolbox has been released.

The HGF Toolbox implements the hierarchical Gaussian filter (HGF) introduced in

Mathys C, Daunizeau J, Friston KJ, Stephan KE (2011). A Bayesian foundation for individual learning under uncertainty. Front. Hum. Neurosci. 5:39. doi: 10.3389/fnhum.2011.00039

After downloading, unzip and read the introduction in the README file.

This release introduces Bayesian parameter averaging and contains bugfixes.

The most important bug concerns the calculation of the negative free energy F, which dropped constant terms in the log-joint for estimation purposes without adding them back in. This does not affect parameter and trajectory estimation. However, F values returned by previous versions of the toolbox are not valid. The only case where this is relevant is where these values were used outside the toolbox for model comparison.


Release Notes
- Introduced Bayesian parameter averaging
- Amended calculation of log-priors in fitModel.m
- Debugged construction of time axis in hgf_plotTraj
- Debugged removal of placeholder field in estimate structure