Título: Predicting chaotic dynamics with reservoir computing
Ponente: Ulrich Parlitz (Max Planck Institute for Dynamics and Self-Organization)
Fecha y hora: 04/12/2024, 12:00
Inscripción online (cierre 30 minutos antes del inicio): https://forms.gle/dvaxftX815mttHvN9
Lugar: Sala de Seminarios del Edificio Torretamarit (CIO) y online
Organizador: José María Amigó García
Abstract:
Reservoir computing utilizes the response of driven dynamical systems for predicting and analyzing time series. After a brief introduction to the basics and classical examples of reservoir computing, we will discuss the role of generalized synchronization and present two approaches to increase the prediction performance. One is the use of delayed values of input and state variables of the reservoir system and the other is the use of parallel reservoirs in combination with linear dimensionality reduction to predict time series of spatially extended, chaotic systems (e.g. Kuramoto-Sivashinsky equation).
Referencia:
Referencias. Parlitz U (2024) Learning from the past: reservoir computing using delayed variables. Front. Appl. Math. Stat. 10:1221051. doi: 10.3389/fams.2024.1221051