Fading memory echo state networks are universal

2020/10/22 by Gonon, Lukas, Ortega, Juan-Pablo · 9 citations
#FOS: Computer and information sciences #Neural and Evolutionary Computing (cs.NE)

paper · doi:10.48550/arxiv.2010.12047

Abstract

Echo state networks (ESNs) have been recently proved to be universal approximants for input/output systems with respect to various L ^p-type criteria. When 1≤ p< ∞, only p-integrability hypotheses need to be imposed, while in the case p=∞ a uniform boundedness hypotheses on the inputs is required. This note shows that, in the last case, a universal family of ESNs can be constructed that contains exclusively elements that have the echo state and the fading memory properties. This conclusion could not be drawn with the results and methods available so far in the literature.

Cited by

• Learning the climate of dynamical systems with state-space systems
• Beyond Lipschitz Continuity and Monotonicity: Fractal and Chaotic Activation Functions in Echo State Networks
• Illuminating the Black Box of Reservoir Computing
• Stochastic dynamics learning with state-space systems
• Feedback-driven recurrent quantum neural network universality
• Organizational Regularities in Recurrent Neural Networks
• A tensor network approach for chaotic time series prediction
• Role of scrambling and noise in temporal information processing with quantum systems
• Learning Dynamical Systems with the Spectral Exterior Calculus

Related

• Echo state networks are universal
• Echo State Networks as State-Space Models: A Systems Perspective
• Empirical Analysis of the Necessary and Sufficient Conditions of the Echo State Property
• Contraction, Criticality, and Capacity: A Dynamical-Systems Perspective on Echo-State Networks
• Towards a Calculus of Echo State Networks