Skip to content

Latest commit

 

History

History
112 lines (89 loc) · 8.98 KB

about.md

File metadata and controls

112 lines (89 loc) · 8.98 KB
Raw
layout title permalink
page
About
/about/

I'm a graduate research intern at the Donders Centre for Cognition. I'm also working at the Data Science Department at Radboud University. At the moment I'm enrolled in a double degree in Artificial Intelligence and Cognitive Neuroscience focusing on machine learning and reinforcement learning. I'm passionate about world model learning, probabilistic deep learning, information theory and neural compression.

Research Interest

My overall interest lies in unsupervised representation learning and its link to intelligent machines. Those interests are currently best encompassed in the fields of model-based reinforcement learning, neural compression and probabilistic deep learning.

  • Model-based Reinforcement Learning: Intrinsically, curiosity driven agents that build a world model within a complex environment fascinate me. Especially agent goal functions that either make use of active inference or information theoretic approaches. I'm particularly intrigued by the problem of long horizon planning which is bottlenecked by the quality of the environment compression.

  • Neural compression: The compression of spatio-temporal dynamics using neural networks for me is one of the most fascinating problems. I would does like to work on the improvement of dynamic variational autoencoders or other deep generative video models. Other possible avenues include the use of attentional mechanisms to enable more and more general model building algorithms (see Perceiver from Deepmind for example). Such models have in my opinion great potential within model-based RL but also in terms of reducing world wide internet traffic.

  • Probabilistic Deep Learning: This topic in my opinion underwrites many of the recent advancements in artificial intelligence and is thus worth to be investigated. I'm especially interested in enabling Bayesian inference on large datasets using variational methods.

Education

  • Radboud University - Nijmegen, Netherlands - Master in Artificial Intelligence (2019 - 2022)

    • Link to the programm
    • specialisation in Cognitive Computing
    • courses like: statistical machine learning, neural information processing and probabilistic deep learning
    • advanced programming courses in Python, Matlab and C++
    • thesis on model-based reinforcement learning with Marcel van Gerven and Danijar Hafner
    • double degree with Cognitive Neuroscience

  • Donders Graduate School - Nijmegen, Netherlands - Research Master in Cognitive Neuroscience (2019 - 2022)

    • Link to the programm
    • specialisation in Neural Computation and Neurotechnology
    • courses like: adv. computational neuroscience, reinforcement learning and deep learning
    • large portion of courses from the neurophysics master
    • double degree with Artificial Intelligence

  • Leiden University - Leiden, Netherlands - Guest Student (2020 - 2021)

    • course on Reinforcement Learning (grade 9)
    • working on MCTS methods and Dreamer

  • Ubaya University - Surabaya, Indonesia - Semester Abroad (2018 - 2019)

    • average grade A (A-D)
    • strengthening intercultural competence
    • course on computer organization and architecture

  • Maastricht University - Maastricht, Netherlands - Bachelor of Science in Psychology (2016 - 2019)

    • cum laude, average grade 8.1(6-10)
    • focusing on motor control and deep learning
    • part of the MaRBLe excellence programme
    • bachelor thesis on Multi-Source Domain Adaptation

  • Goethe University - Frankfurt, Germany - Bachelor of Science in Economics (2013 - 2016)

    • average grade 1.9 (1.0-4.0)
    • focusing on micro- and macroeconomics (game theory, cooperation, agency theory)
    • courses like: business cycle theory and policy, labor economics, finance and inequality
    • bachelor thesis based in empircal research about cooperation norms within an Public Good Game

Research Experience

  • Graduate Research Intern at the Donders Institute for Brain, Cognition and Behaviour (2021 - Present)

    • working on model-based reinforcement learning with a special, improving agent internal goal functions to facilitate long horizon planning
    • with Marcel van Gerven and Danijar Hafner

  • Student Assistant at the Data Science Department of Radboud University (2021 - Present)

    • working on AutoMATE project, a framework designed for procedural assignment generation with seamless autograding integration
    • working with Ioan Gabriel Bucur
    • focusing on the courses statistical machine learning and data mining

  • Lab roation at the Donders Institue for Brain, Cognition and Behaviour (2018 - 2019)

  • Bachelor thesis research at Maastricht University (2018 - 2019)

    • part of the research based excellence program partially organized by EDLAB
    • working on convolutional neural networks optimized for multi-modal scene recognition
    • re-using a network pretrained on the places205 dataset (natural images) for clip art and sketches
    • implemented using Tensorflow

  • Research Assistant at Mainz University (2014)

    • determining determinants of cooperation in public good games
    • in cooperation with the Faculty of Business and Economics Management and Microeconomics at Goethe-University
    • planning, coordination, documentation and analysis of the results

Projects

  • Master Thesis (ongoing) (Donders Institut and Radboud University) [Blog]
  • AutoMATE (ongoing) (Data Science Department, Radboud University)
  • World Model Agents (Radboud University, Natural Computing course) [Blog]
  • Active Inference Notebook (Radboud University, Lab rotation) [GitHub]
  • Neuromorphic Active Inference Agents (Radboud University, Neuromorphic Computing course) [Blog]
  • A3C Agents (Radboud University, Neural Information Processing course and Lab rotation) [Blog]
  • Multi-Source Domain Adaptation (Maastricht University, Bachelor thesis) [Paper / Poster / Blog]

Teaching Experience

Skills

  • Programming: Python (numpy, seaborn, scikit, jupyter, etc.), R, Matlab, HTML. Basics in Julia and C++
  • Deep learning frameworks: Tensorflow, Pytorch and Keras
  • Experience with AWS, Deepnote, Google Colab and Google Cloud Platform
  • Experienced in using Git, LaTeX and Linux