Nonlinear system analysis #

Coursework for my class on nonlinear dynamical systems & chaos. Read the report and the Jupyter notebooks on GitHub, or, for a quick overview, check out some captioned figures in the ReadMe.

Master's thesis: machine learning for neuroscience researchers #

I had the opportunity to do my engineering master’s thesis in the lab of Fabian Kloosterman at NERF, where I was introduced to the wonderful world of neuroscience. The context and the goals of my thesis, and the current results, are shortly explained in this conference abstract. This poster, made at the beginning of the project, visually shows the setup and the initial plan. I developed a detection algorithm based on a gated recurrent neural network, yielding a significant improvement over the state of the art. (For the hippocampus connoisseurs: here are some preliminary detection results. Blue: state of the art method. Magenta: new method).

Electives I took in my master's #

Machine learning and data analysis:

• Bayesian & probabilistic-graphical models,
• Artificial neural networks,
• ML 101, ML in practice,
• Multivariate & robust statistics,
• Bio-informatics & more Bayesian modelling, and
• Nonlinear systems.

Neuroscience:

• Methods in systems neuroscience 101,
• Computational neuroscience 101.

Misc and from core curriculum:

• IT security
• Human physiology, Cell biology
• Anatomy, Biomechanics, Bio-fluid mechanics
• Sensors & stimulation in healthcare
• Biomaterials, Design in medical tech, Medical imaging

Web interface to label neuro data #

The signals I had to detect for my master’s thesis were not clearly defined – there is no ground truth for them. I therefore made a web interface for the neuroscientists in the lab, to label potential events. Tech: Vue.js, Node.js, mongodb, Bootstrap, Heroku, and the scientific Python stack for data processing and plot generation. GitHub repo’s for frontend, and backend.

⚠️ NOTE: the free service on which the backend was hosted has gone bankrupt; so the full interface won’t load anymore, alas.

Interactive 3D rat brain #

I wanted to learn about the gross neuroanatomy of the animals in the lab. Learning brain anatomy from books is not very useful though. Inspired by the great Allen Institute Brain Explorer for mice, I therefore made this interactive 3D visualisation of structures in the rat brain, based on open data. The GitHub repo explains how this data was processed. (Note: check out 3dBAR for another 3D rat brain atlas, with more features).

3D-printed rat hippocampus #

We have a Formlabs stereolithography 3D printer in the lab. So why not print some of the structures from the above 3D rat brain?

Data science internship: wearable motion sensor validation #

Byteflies makes medical wearables – small devices full of sensors that can be worn anywhere on the body. My job was to validate the signal quality of the motion sensors (accelerometer, gyroscope, and magnetometer) against the golden standard: motion capture. Read all about it in this blog post.

Software engineering internship: usage data pipeline #

DataCamp makes online courses to learn ‘data science’ skills, via videos and interactive exercises. The DataCamp team wanted to know how long users spend in each exercise, and to have an internal API available that exposes this data, grouped and sliced by different dimensions. So that’s what I made, with the above high-level architecture.

Poster design #

I did the poster typography and art selection for concerts with my symphony orchestra at the time (the lovely Arenbergorkest). For this orchestra I engaged myself in different roles: treasurer, IT, second violins section leader, and president during our concert tour in China.

Poster design, science edition #

First poster made at the beginning of the master thesis, for the yearly NERF Retreat. Second poster made near the end of the thesis, for the Cell-NERF Symposium on Neurotechnologies.

Programming memories #

I loved programming as a kid (still do and still am). Some of the hobby projects I remember:

• Gravity & planetary motion simulator. The generated trajectories were interesting. Programmed, as many other projects, in Flash / ActionScript (RIP this beautiful programming environment).
• Weierstrass function plotter. I was fascinated by this function that was continuous everywhere, but nowhere differentiable. With this app you could zoom in infinitely close on the function (or at least until the edge of numerical floating point precision).
• Fractal tree generator (including a clumsily animated giraffe roaming between the trees).
• PID controller for a two-wheeled balancing robot (like a Segway) made with Lego Mindstorms.
• A chess AI. (Prototype in Python, final optimised code in C, using vectorized CPU instructions).
• A Scrabble AI (directly interfacing with the Wordfeud API).