MSc thesis topics
Topic proposals for 2023-2024 can be found below. Please check regularly for updates.
Personal research project in deep learning
This thesis proposal is not tied to a specific project. Instead, it welcomes students to make topic proposals on open research problems of their choosing and interest. Proposals should be centered around deep learning. Examples of projects include:
- Theoretical research in deep learning
- Improvements to existing deep learning algorithms or models
- Application of deep learning to solve a real-world problem
- Development of deep learning software
Finding a research problem to work on is considered as part of this thesis subject. Students should come with a concrete and well-defined thesis topic. Proposals will be reviewed and discussed with the student before their approval (if any). [PDF]
Contact: Gilles Louppe.
Simulation-based inference in natural sciences
In many areas of science, complicated real-world phenomena are best described through computer simulations. Typically, the simulators implement a stochastic generative process in the forward mode based on a well-motivated mechanistic model with parameters \(\theta\). While the simulators can generate samples of observations \(x \sim p(x | \theta)\), they typically do not admit a tractable likelihood (or density) \(p(x | \theta)\). Probabilistic models defined only via the samples they produce are often called implicit models. Implicit models lead to intractable inverse problems, which is a barrier for statistical inference of the parameters \(\theta\) given observed data.
Depending on the student interests, the goal of this project will be to apply simulation-based inference algorithms developed within our research group to a scientific problem of her choosing. Direct collaboration opportunities include projects with climatologists, geologists, particle physicists or astronomers. Methodological improvements of those algorithms are also possible and welcome within the scope of this thesis.
Contact: Gilles Louppe.
Simulation-based inference for exoplanet astrometry
Detection and characterization of exoplanets using direct imaging is one of the most exciting frontiers in modern astronomy. By capturing a series of snapshot images of an exoplanet, we can understand the properties of its planetary system. Fitting an exoplanet’s astrometry data allows us to understand the planet’s formation and evolution, and sometimes even detect unseen planets. A planet’s orbit is parameterized by eight Keplerian elements: semi-major axis, eccentricity, inclination angle, argument of periastron of the secondary’s orbit, longitude of ascending node, epoch of periastron passage, parallax, and total mass. In an orbit fitting problem, we estimate the posterior of these parameters based on the exoplanet astrometry data relative to the primary star from telescope snapshot images. Due to the large amount of observational data, the high dimensionality of features to infer, and the potential multi-modality of posterior distributions, recovering a full posterior distribution is often computationally challenging. Traditionally, sampling-based approaches, such as importance sampling and Markov Chain Monte Carlo (MCMC) methods, are widely used to solve inference problems in computational science. But without a good proposal distribution in importance sampling or a good random initialization in MCMC, sampling approaches are often prohibitively slow to converge.
In this project, we propose to use some of the latest developments in the field of simulation-based inference to speed up the recovery of orbital parameters from exoplanet imaging data sets, without making any simplifying assumption on the shape of the posterior distribution for these parameters. We will aim to reproduce the results of Sun et al (2022) but using generic inference algorithms such as neural ratio estimation, neural posterior estimation, or neural score estimation.
Contact: Gilles Louppe, Olivier Absil.
(Reproduced from Sun et al, 2022)
Graph neural networks for models of active matter
Many collective systems exist in nature far from equilibrium, ranging from cellular sheets up to flocks of birds. These systems reflect a form of active matter, whereby individual material components have internal energy. Under specific parameter regimes, these active systems undergo phase transitions whereby small fluctuations of single components can lead to global changes to the rheology of the system. Simulations and methods from statistical physics are typically used to understand and predict these phase transitions for real-world observations.
The goal of this project will be to investigate the use of graph neural networks to model the dynamics of active matter systems. Depending on the student interests, the project will focus on either the generation of synthetic data or the inference of model parameters from synthetic and real-world data.
Contact: Gilles Louppe.
Videos of collective cell migration observed in vitro (data collected by Dr. Namid Stillman). The goal of this project could involve (i) learning a graph-based generative model of this data, or (ii) the application and development of graph-based simulation-based inference on biophysical models that can explain these observations.
Simulation-based inference of neural models from spikes
A fundamental question in neuroscience is how to link observed neural activity to the unobserved biophysical mechanisms that generate this activity. Therefore, there is a critical need for methods to incorporate the partial and noisy data that we observe with detailed, mechanistic models of neural activity.
In this project, we will explore how to estimate the parameters and the hidden variables of neuronal models from neuronal spike train responses. In particular, we will compare modern simulation-based inference methods to more traditional methods like particle filters. Depending on the progress, we will also investigate how to actively collect new data in closed-loop experiments to improve the inference. [PDF]
Contact: Gilles Louppe, Pierre Sacré.
Deep learning for sport videos [EVS Broadcast Equipment]
In collaboration with EVS Broadcast Equipment, several master thesis topics are proposed on deep learning for sport videos. Topics include photo-realistic data generation, diffusion models, video frame interpolation, novel view synthesis, speech recognition, or automatic tracking, among others. The exhaustive list of projects is available in the document below. [PDF]
Contact: Gilles Louppe, Oliver Barnich.
Data science in water management [Purecontrol]
In collaboration with Purecontrol, several master thesis topics are proposed on data science in water management. Topics include the prediction of phosphorus concentration in wastewater treatment plants, the development of algorithms for anomaly detection in these plants, and the modeling of the environmental effect caused by water flows. The exhaustive list of projects is available in the document below. [PDF]
Previously supervised MSc thesis (2018-Present) can be found on Matheo.