Title: Forecasting and the universality problem in dynamic machine learning
Authors: Lyudmila Grigoryeva - University of Konstanz (Germany)
Juan-Pablo Ortega - University St. Gallen (Switzerland) [presenting]
Abstract: A relatively recent family of dynamic machine learning paradigms known collectively as reservoir computing is presented which is capable of unprecedented performances in the forecasting of deterministic and stochastic processes. We will then focus on the universal approximation properties the most widely used families of reservoir computers in applications. These results are a much awaited generalization to the dynamic context of previous well-known static results obtained in the context of neural networks.