Title: Machine learning with universal reservoir computers using non-homogeneous state-affine systems and forecasting tasks
Authors: Lyudmila Grigoryeva - University of Konstanz (Germany) [presenting]
Juan-Pablo Ortega - University St. Gallen (Switzerland)
Abstract: A new class of non-homogeneous state-affine systems is proposed which can be used for various machine learning applications. Sufficient conditions are identified that guarantee first, that the associated reservoir computers with linear readouts are causal, time-invariant, and satisfy the fading memory property and second, that a subset of this class is universal in the category of fading memory filters with stochastic bounded inputs. This means that any discrete-time filter that satisfies the fading memory property with random inputs of that type can be uniformly approximated by elements in the non-homogeneous state-affine family. We empirically demonstrate the competitive performance of the proposed non-homogeneous state-affine systems in time series forecasting tasks.