Asymmetrically connected reservoir networks learn better

Abstract

We show that connectivity within the high-dimensional recurrent layer of a reservoir network is crucial for its performance. To this end, we systematically investigate the impact of network connectivity on its performance, i.e., we examine the symmetry and structure of the reservoir in relation to its computational power. Reservoirs with random and asymmetric connections are found to perform better for an exemplary Mackey-Glass time series than all structured reservoirs, including biologically inspired connectivities, such as small-world topologies. This result is quantified by the information processing capacity of the different network topologies which becomes highest for asymmetric and randomly connected networks.

6 pages, 4 figures, supplementary material

References in corpus (10)

• Modularity and community structure in networks
• Using Machine Learning to Replicate Chaotic Attractors and Calculate Lyapunov Exponents from Data
• Network Structure Effects in Reservoir Computers
• Hybrid quantum-classical reservoir computing of thermal convection flow
• Reservoir computing model of two-dimensional turbulent convection
• Efficient implementations of echo state network cross-validation
• Chaotic attractor reconstruction using small reservoirs -- the influence of topology
• Generalizability of reservoir computing for flux-driven two-dimensional convection
• Revealing directed effective connectivity of cortical neuronal networks from measurements
• The numerical computation of unstable manifolds for infinite dimensional dynamical systems by embedding techniques

Cited by in corpus (1)

• Prediction performance of random reservoirs with different topology for nonlinear dynamical systems with different number of degrees of freedom