I propose a few topics for bachelor and master theses related to the recent research interests in our research group (the list may be updated soon). Please contact me for more details.

[INZ/LIC/ISIM] Uncertainty in Time Series Embeddings for Financial Time Series

This topic concerns vector representations of time series, so-called time series embeddings, usually generated by deep learning or representation learning approaches, such as TS2Vec [PDF], TRep [PDF], etc., and aims at studying their resistance for data uncertainty [PDF], especially in the context of financial time series. I suggest to focus on uncertainty in time series of limit order books [PDF] from the London Stock Exchange (LSE ROB) and to start from a few regular approaches with available prototype implementations. Please contact me for more details.

[INZ/LIC/ISIM] Generative AI and Diffusion Models in Time Series Embeddings for Financial Time Series

This topic concerns generative AI, especially diffusion models [PDF], for time series [PDF] and their influence on internal vector representations of time series in deep learning or representation learning approaches, such as TS2Vec [PDF], TRep [PDF], etc. It aims at improving time series forecasting, especially in the context of financial time series, such as limit order books [PDF] from the London Stock Exchange (LSE ROB). Please contact me for more details.

[INZ/LIC/ISIM] Generative AI in Deep Learning for Dynamic Graphs

This topic concerns generative AI, especially diffusion models [PDF], for time series [PDF] and their application in deep learning approaches for dynamic graphs, such as TGAT [PDF] or TGN [PDF]. It aims at improving forecasting of events in such dynamic graphs with generating additional data and making the model more accurate. Please contact me for more details.