News:

Information:

This year, due to some possible absences related to research visits, our meetings will start in the second half of the semester, but there will be two meetings per week (as noted in the enrollment system).

Organization, Rules, Topics:

This project will focus on machine learning and deep learning techniques for time series and temporal data. I assume that we will work together on a larger general project divided into parts assigned to particular teams (some parts may be larger, some smaller, teams may change in time, if needed; we will work in a type of the agile methodology). We will meet on a weekly basis to discuss the progress, address the emerging problems, and learn new knowledge. I suppose that the work will not only be programming our own solutions, but also (or perhaps above all) reading and exploring efficient solutions published recently in top conferences or journals. However, the tasks will be different and will depend on the competences and preferences of individual participants.

I propose to determine the exact scope and the exact goal of the project on the first meeting, taking into consideration the competences and preferences of the participants, but I assume that some of the following techniques might be useful in the project:
- time series representation learning (e.g. T-Loss, TST, TNC, TS2Vec, TRep),
- time series augmentation (e.g. with diffusion models),
- time series segmentation with state space models (e.g. HMM, SLDS, Kalman Filters),
- anomaly detection in spatio-temporal data,
- missing values imputation in spatio-temporal data,
- satellite image time series forecasting,
- human activity recognition from spatio-temporal data,
- etc.