I conduct research on general machine learning methodology and find its applications in healthcare/neuroscience. Back at Jacobs, I first worked with Prof. Michael Kohlhase and then my real ML research started with Prof. Herbert Jaeger and Prof. Benn Godde. I also spent time at MPI with Prof. Grosse-Wentrup and at EPFL with Prof. Courtine on brain-computer-interface research which was super interesting. My current master thesis supervisors are Dr. François Fleuret and Dr. Mathieu Salzmann.
|2019 Jun: A Primer on the Delayed Adversarial Attack in Using Recurrent Neural Networks for Reinforcement Learning
An introduction of a series of work on delayed adversarial attack in deep learning. We will make the code available soon.
|Repo||2019 Feb: Reinforcement Learning Cheat Sheats
A collection of theories and applications of reinforcement learning as part of my master thesis.
|Repo PDF||2017 Jun: Motor Learning Skill Classification Using Echo State Networks
ESNs are a type of Recurrent Neural Networks (RNNs) that are easy to train but still maintain the power of capturing the temporal information that do. In my bachelor thesis, I constructed ESNs for motor learning skill classification on EEG data. Additionally, I also developed new methods on how to increase the interpabiltiy of ESNs.
Mathematical Knowledge Management (MKM)
|2016 Jun: Mixing Surface Languages for OMDoc
The formalism of mathmatics: MathML, OpenMath and OMDoc are machine-friendly formats for math so that the machines can do search, automated proving and cross-referencing easily. TeX is the main driving horse for high quality mathematics documents authoring. A variant of TeX that allows for structural formats, enables us to transform those documents into OMDoc with the help of LaTeXML sTeX plugin to let machines operate on the documents.