Probabilistic Modeling

This cluster consists of Ph.D. students from a variety of universities around Sweden who are interested in topics related to Probabilistic modeling. The core theme of the cluster is modeling, approximating and understanding different phenomena through the formalism of probabilism. Common interests among cluster members include studies of state-space models, Bayesian inference, energy-based methods, normalizing flows, stochastic (partial) differential equations, computer vision, probabilistic weather forecasting, neuro symbolic artificial intelligence, nonlinear filtering, simultaneous localization and mapping, deep learning convergence theory and many more. This research can be applied in fields such as telecommunication, radar technology, pharmaceutical applications, control systems, autonomous systems, etc.

Cluster activities

Currently the cluster gathers digitally once a month. The meetings allow for each member to give a 5-10 minute update on his/her current research or what’s happened since last time. The meetings further facilitate spreading of general information, which could include tips on courses or updates from WASP. The meetings are concluded with a slightly longer/more in-depth presentation from one of the members.

The cluster plan for 2024 is to initiate regular physical meetings as well as a reading group (separate from the monthly meetings).
Each cluster member is free to participate in as much or as little as he/she wants with little pressure. Note however that engagement drives more engagement which results in a thriving cluster for all members.

Joining the cluster

If you are interested in joining the cluster or getting into contact, feel free to contact the cluster leader, Anton.

Members of the cluster

Kasper Bågmark

Kasper Bågmark
  • Affiliation: WASP Ph.D. student at the department of Mathematical Sciences, Chalmers
  • Main supervisor: Annika Lang
  • Research:
    • My research is concerned with applying deep learning to solve the optimal filtering problem in a nonlinear setting. Current state of the art algorithms such as particle filters suffers from the curse of dimensionality and the main objective is to avoid this by achieving reasonable scaling. Filtering consist of finding the conditional probability density of a hidden state given observations that are affected by noise. Currently I am working with the Zakai equation which is a stochastic partial differential equation which the filtering density solves.
  • Keywords: Bayesian statistics, stochastic partial differential equations, filtering problem

Gizem Çaylak

Gizem Çaylak
  • Affiliation: WASP Ph.D. student at KTH
  • Main supervisor: David Broman
  • Research:
    • My research focuses on deep probabilistic programming languages and their application on clinical data.
  • Contact: caylak@kth.se

Hedieh Khosravi
  • Affiliation: WASP Ph.D. student at Lund University
  • Main supervisor: Fredrik Tufvesson
  • Research:
    • My research interests focus on Wireless Communications in particular mmWave channels and radio based positioning. The positioning capability is a cornerstone for additional services in 5G wireless systems. In order to really take the advantage of 5G in, e.g., industrial automation, the communication service should be combined with a highly accurate positioning service. We investigate how to achieve this in practice by using the already available communication signals.

Anton Kullberg
  • Affiliation: WASP Ph.D. student at the Division of Automatic Control, Department of Electrical Engineering, Linköping University
  • Main supervisor: Gustaf Hendeby
  • Research:
    • Concerned with real-time joint state estimation and model learning/system identification. Particularly, I am interested in bridging the gap between classical sensor fusion techniques such as the Kalman filter, with the machine learning/system identification domain to enable Kalman filter like algorithms for learning (partially) unknown system dynamics online. Currently working with general basis function expansions interpretable as parametrizations of Gaussian processes.
  • Contact: anton.kullberg@liu.se
  • Affiliation: WASP Ph.D. student at the Computer Vision Group, E2 Department, Chalmers
  • Main supervisor: Christopher Zach
  • Research:
    • Topics include 3D representation learning and energy-based models.
  • Contact: xixil@chalmers.se
  • Affiliation: WASP Ph.D. student at the Computer Vision Group, E2 Department, Chalmers
  • Main supervisor: Christopher Zach
  • Research:
    • Learning and Leveraging Rich Priors for Factorization Problems: combining traditional parametric formulations induced by domain expertise with non-parametric models learned from examples; main application of interest are factorization-based problems, in particular non-rigid structure-from-motion.
  • Keywords: Geometric computer vision, Representation learning, 3D vision
  • Affiliation: WASP Ph.D. student at Department of Computer and Information Science, Linköping University
  • Main supervisor: Fredrik Lindsten
  • Research:
    • Reliability of deep learning models, including uncertainty estimation in deep neural networks and weak supervision.

Joel Oskarsson
  • Affiliation: WASP affiliated Ph.D. student at the Division of Statistics and Machine Learning, Department of Computer and Information Science, Linköping University
  • Main supervisor: Fredrik Lindsten
  • Research:
    • Machine learning methods for data with spatial-, temporal- and graph-structure, including combinations of these. Bridging probabilistic, Bayesian and deep learning for structured data.
  • Contact: joel.oskarsson@liu.se

Cluster meetings & milestones

All meetings include general information and discussions. In the list below the particular focus of that meeting is listed.

2021:

  • February: Introduction meeting
  • March: Kasper presented his research
  • April: Anton presented his research
  • May: Daniel presented his research
  • June: Gizem presented her research
  • September: Amanda presented her research
  • October: Alexander presented his research
  • November: Yaroslava presented her research

2022:

  • January: Winter conference, we discussed our posters and some new members were recruited
  • February: Xixi presented her research
  • March: Kasper presented his research
  • June: Rajmund Nagy guested the cluster together with people from the generative models cluster talking about probabilistic diffusion models
  • September: Marcus presented his research
  • November: Joel presented his research

2023:

  • January: Winter conference, over 30 participants and discussions about the future of the cluster were held
  • September/October: Cluster kick-start after long period of hibernation (due to change in cluster leadership)
  • November: Monthly meeting – Anton presented a paper.
  • December: Monthly meeting – Joel talked about his research visit in London (and probabilistic weather forecasting).

2024:

  • January: Winter conference – about 20 participants got to know each other and talked about cluster expectations for 2024.
  • February: Monthly meeting.
  • February: Name change from Bayesian Statistics, SDEs and Probabilistic Programming to Probabilistic Modeling to align with cluster interests.
  • March: Monthly meeting – Amanda talked about news in AI.
  • April: Monthly meeting.
  • May: Monthly meeting.

Contact

Anton Kullberg

Academic PhD student AS batch 3, Division of Automatic Control, Linköping University

Send Email