1. I am attending Bayes@Lund on 20 April 2017 and will be giving a talk on eye movement analysis using Gaussian Process. I am mostly experimenting in GPflow and PyMC3. You can find the slides and materials on Github.
3. I updated my PyMC3 port of Lee and Wagenmakers' Bayesian Cognitive Modeling following the recent updates in PyMC3. I am quite pleased with the current version, as all models are running properly now after vectorization (putting my Matlab skill into used). There are so many exciting new features in PyMC3, and the development community is just fantastic.
4. I updated my Gist on Testing hypotheses via a mixture estimation model. In the previous experiments with this approach, I often discover that the trace is not mixing very well. The main reason is the multimodality of the mixture model. Inspired by Michael Betancourt's case study in Stan, I also sorted the mixture weight to break the multimodality. However, one must be careful of placing the order of the mixture component, as the less possible component should be placed in the front. As a heuristic, we can place the component coding for the null hypothesis before the alternative hypothesis, as our strawman null hypothesis is usually false.