Teaching and supervision - Spring 2024
During the Spring Semester 2024, I will be teaching/supervising:
- 02450 Introduction to machine learning and datamining (course coordinator, lecturing in 9 weeks), DTU
- Six BSc and two MSc projects, DTU (as main supervisor)
- Special courses (deep learning for audio processing, reinforcement learning), DTU
A complete list of standard courses and a list
of representative examples of projects are available below.
Prospective BSc/MSc thesis students
My supervision approach: It is your project, not mine... I meet with students approx. every two weeks for 30-45 min, or as needed (typically more frequently at important/critical periods in the project).
Students can book a time slot (various duration) on Thursday. I will share a link to the booking calendar, once we haev agreed on a project.
At the meetings we dicuss achivements, problems, plans (for the next 1-3 weeks).
Prospective BSc thesis students
Predefined: I am looking for BSc students (typically enrolled at at DTU Compute) to work on the following projects at DTU
in Fall 2024 (the projects will be adjusted based on the educational background of the allocated student(s)). I recommend working in a group of 2-3 people.
- Bayesian optimisation for multi-modal optimisation (based on trust region methods such as https://arxiv.org/pdf/1910.01739.pdf, https://arxiv.org/pdf/2210.10953.pdf).
- Conditional (latent) diffusion models (e.g. for generative modelling of microscopy images or molecules).
- Scaling machine learning to massive graphs (e.g. for protein modelling, e.g. https://arxiv.org/abs/2102.03150).
- Deep learning for speech/audio processing and analysis (e.g. seperation, denoising, classificaiton).
Self-defined projects (e.g. with industry or based on your own dataset/problem): I am open to self-defined projects iff
it aligns with my own interests. I expect you approach me with a concrete proposal and a refelection on how the project relates
to my interests, and why I would be a suitable supervisor.
Your background: I would expect you are enrolled on the Artificial Intelligence and Data study line or
the Mathmatics and Technology study line (combined with a keen interest in machine learning) at
DTU Compute - or have equivalent experience/skills.
Application/contact: Feel free to email me stating your interests and idealy attach a list of completed courses/grades. Please indicate your study line/degree programme
and whether you are more interested in the application (of machine learning) or theory/algorithms.
Prospective MSc thesis students
Predefined: I am looking for MSc students (typically enrolled at at DTU) to work on the following projects
in Fall 2024 (the projects will be adjusted based on the educational background of the allocated student(s)). I recommend considering working in a group of two people.
- Bayesian optimization for multi-modal problems (beyond trust region methods)
- Flow matching (based on https://arxiv.org/abs/2210.02747) for generative modelling (e.g. of molecules or dynamic vector fields)
- Machine learning for molecular simulation (based on https://openreview.net/pdf?id=TnIZfXSFJAh)
- Diffusion modelling/flow matching for speech/audio processing and analysis.
- Deep/machine learning for team sports (with external partners)
Self-defined projects (e.g. with industry or based on your own dataset/problem): I am open to self-defined projects if
it aligns with my own interests. I expect you approach me with a concrete proposal and a refelection on how the project relates
to my interests, and why I would be a suitable supervisor.
Your background: I would expect that you have a solid background in machine learning (e.g. obtained via DTU courses 02450, 02456, 02476, 42186, 02477, 02460).
Some projects require specific skills/experience in image-analysis (e.g. 02516) or maths/stochastic processes (e.g 02407).
Application/contact: Feel free to email me indicating your interests. Ideally include a complete list of courses/grades
and a short reflection on how your specific skills/experience (in particular relating to machine learning, maths and programming) will help
you complete the project you have in mind.
Prospective BEng students (DA: diplompraktik/-projekter)
I do not currently supervise BEng projects/internships.
Completed BSc/MSc/MSci projects - representative examples
I have supervised more than 50 BSc/MSc/MSci projects at the Technical University of Denmark (DTU) and the University of Glasgow (UofG).
A few representative examples for inspiration for prospective students:.
MSc/MSci projects:
2023 Predicting kinetic stability of proteins using machine learning, DTU
2023 Self-supervised learning for audio classification (with Michael Riis as main supervisor), DTU
2022 Investigation of Deep Probabilistic Surrogate Models in Bayesian Optimization (with Mikkel N. Schmidt as main supervisor)
2022 Outlier detection and data imputation in multivariate time series, DTU
2022 Recognizing key patterns in football using positional data [using ML 4 graphs], DTU (with external partners)
2021 A Multi-Task Approach to Hearing Aid Fine-Tuning, R.S, UofG (with external partners)
2021 Evaluating Uncertainty Estimation Methods For Deep Neural Networks In Inverse Reinforcement Learning, J.K, UofG
2020 Representation Learning with Sparse VAEs for Single-Cell Images, A.R., UofG
2020 Tangent propagation for environmental sound classification, B.H., UofG
2018-2019 Variational Inference in Bayesian Inverse Reinforcement Learning, P.D., UofG
2017-2018 Machine learning for understanding human decision-making, P.C., UofG
2018 Accelerating Image-Based Cell Profiling with Machine Learning, L.L, UofG [best UofG SoCS project award]
2017 Perceptual Embeddings of Music Objects using Machine Learning, H. W., UofG
2017 A toolbox for performance modelling and selection of dense neural network architectures, F. B., UofG
BSc / Honours projects:
2024 Scaling properties of graph neural networks for molecular modeling
2024 Deep learning for speech enhancement, DTU (with Kenny Olsen)
2023 Large language models for collecting structured information, DTU (with Finn A. Nielsen)
2023 and 2024 Conditional generative modelling using diffusion models, DTU
2023 Learning audio representations for search and retrieval using variational autoencoders, DTU (with Michael Riis and external partners).
2022 Modelling Predatory Behaviour in Social Chatrooms using AI [i.e. graphs], DTU (with external partners)
2020 Towards Efficient Reasoning with Deep Neural Networks, G.M, UofG
2020 Machine learning based optimisation of parameters in audio pipelines, A.F. UofG
2017-2018 Preference Elicitation for Music using Spoken Dialogue Systems, M.B, UofG
2017-2018 Hybrid Music Recommendation, A.G., UofG
2017-2018 Neural Networks for Dialogue Systems, Z.I, UofG
Taught courses
I have taugth several courses at various levels in the UK (University of Glasgow) and Denmark (Technical University of Denmark).
2024 Spring: 02450 Introduction to machine learning and datamining (with Georgios Arvanitidis as co-lecturer), DTU
2023 Fall: 02450 Introduction to machine learning and datamining (3 weeks, led by Georgios Arvanitidis ), DTU
2023 Fall: 26255 Molecular dynamics and machine learning (2 weeks, led by Günther H.J. Peters ), DTU
2023 Spring: 02450 Introduction to machine learning and datamining (with Jes Frellsen as co-lecturer), DTU
2022 Fall: 26255 Molecular dynamics and machine learning (2 weeks, led by Günther H.J. Peters ), DTU
2022 Fall: 02456 Deep learning (project supervisor, led by Ole Winther and Jes Frellsen), DTU
2022 Fall: 02450 Introduction to machine learning and datamining (3 lectures, led by Georgios Arvanitidis ), DTU
2022 Spring: 02450 Introduction to machine learning and datamining (with Jes Frellsen as co-lecturer), DTU
2021-2022: Artificial Intelligence COMPSCI4004 H (Semester 1, with Dr Jonathan Grizou), UofG
2020-2021: Artificial Intelligence COMPSCI4004 H (Semester 1), UofG
2019-2020: Introduction to Data Science and System COMPSCI 5087 (M) (Semester 1), with Dr Jeff Dalton and Dr Nikos Ntarmos, UofG
2019-2020: Artificial Intelligence COMPSCI4004 / COMPCI5987 M/H (Semester 1), UofG
2018-2019: Artificial Intelligence COMPSCI4004 (Semester 2), UofG
2018-2019: Music Curation and Analysis (H) ARTMED4038 (Semester 1, three weeks), with Dr Tim Duguid, UofG
2018-2019: Software Engineering (M) COMPSCI5059 (Semester 2), with Dr Ke Yuan, UofG
2017-2018: Artificial Intelligence COMPSCI4004 (Semester 1), UofG
2017-2018: Software Engineering (M) COMPSCI5059 (Semester 2), with Dr Ke Yuan, UofG
2016-2017: Artificial Intelligence M/H COMPSCI4004 (Semester 1), UofG