Description

Variational Autoencoders (VAE) are deep learning models with diverse applications in various fields. Students will learn and apply VAEs to generate and predict financial time series. They will also explore how VAEs can be applied in time series classification and parameter estimation problems. The project will use Python with PyTorch as a machine learning platform.

Supervisors

• László Varga - Citi Markets Quantitative Analysis and ELTE TTK, Department of Probability Theory and Statistics, laszlo.varga@ttk.elte.hu
• Gábor Fáth - ELTE RiskLab and ELTE TTK, Department of Physics of Complex Systems, gabor.fath@ttk.elte.hu

Prerequisites

• Language: English
• Programming: Python
• Basic knowledge of machine learning, neural networks, time series

References

• D. P. Kingma, M. Welling (2019): An Introduction to Variational Autoencoders. Foundations and Trends in Machine Learning
• A. Desai, C. Freeman, Z. Wang, I. Beaver (2021): Timevae: A variational auto-encoder for multivariate time series generation, arXiv
• H. I. Fawaz, G. Forestier, J. Weber, L. Idoumghar, and P.-A. Muller (2019): Deep learning for time series classification: a review, Data mining and knowledge discovery, Springer
• Pytorch guide: https://pytorch.org/tutorials/