Deep Learning for Time Series: Advancements in Recurrent Architectures and Reservoir Computing
Associate professor of Deep Learning at the Department of Computer Science of the University of Pisa.
My research interests are at the intersection between the areas of Machine Learning, Deep Learning, Neural Networks and Dynamical Systems.
Specifically, I am interested in Dynamical Neural Network models, Reservoir Computing, and Deep Learning for Graphs.
This seminar examines core advancements in deep learning for time series, spanning Recurrent Neural Networks, Transformers, and state space models such as MAMBA and S4. It further addresses Reservoir Computing, with a focus on new architectures suited for non-dissipative dynamics, essential for maintaining long-term temporal information and go beyond the classical limitations of conventional reservoirs. The discussion will contextualize these models within theoretical frameworks and practical implications for robust time series analysis.
Link to talk:
https://us06web.zoom.us/j/87108161264?pwd=iyNbExZYut49VbasIw7Cym1KiYWm7Y.1
Contact details:
Miguel C. Soriano Contact form