Jesper Wohlert
[ ] am a machine learning engineer at Special Project group working on autonomous systems software.
In the past I've developed machine-learning based planning algorithms using reinforcement learning and trajectory optimisation for cyber-physical systems. These systems are currently deployed in real-time safety critical environments such as chemical processing plants and autonomous mining vehicles.
Some of my interests in machine learning include representation learning, i.e. what consistutes a good representation for a given problem and how to learn it. I am also interested in how learned systems can be validated and interpreted such that one can give reason about the system's function.
Research
We present a method for simultaneous density estimation and manifold learning without the
need for prescribing geometric priors.
Open source
Implementations of semi-supervised deep generative models that were popular around 2018.
Implementation of the generative query network paper by DeepMind (Eslami et al).