A visualization as well as a tutorial on reinforcement learning algorithms, starting with the good old gradient descent and ending with the policy gradients algorithms.
The algorithms run in the browser using Tensorflow.js and are visualized using pure D3.js.

A case study on the applicability of solving reinforcement learning (RL) tasks when posed as an inference problem.

A review of semi-supervised learning methods using GANs. A prevalent theme in these methods was simply combining a supervised classification loss with the unsupervised GAN loss in order to make use of the huge set of unlabeled data.

Metropolis within Gibbs and HMC inference engines for a first-order probabilistic programming language (FOPPL) compiler as well as a likelihood weighting interpreter for a higher-order probabilistic programming language (HOPPL).

Simulating snake locomotion using PyBullet.

A simple simulation of a multi-link pendulum.

Compiler for a restricted version of the Java language, written in C++.