Symmetry-Aware Reservoir Computing

Authors:
  • WBWendson A. S. Barbosa
    Wendson A. S. Barbosa
  • AGAaron Griffith
    Aaron Griffith
  • GRGraham E. Rowlands
    Graham E. Rowlands
  • LGLuke C. G. Govia
    Luke C. G. Govia
  • GRGuilhem J. Ribeill
    Guilhem J. Ribeill
  • MNMinh-Hai Nguyen
    Minh-Hai Nguyen
  • TOThomas A. Ohki
    Thomas A. Ohki
  • DGDaniel J. Gauthier
    Daniel J. Gauthier

Abstract
We match the symmetry properties of a reservoir computer (RC) to the data being processed dramatically increasing its processing power. We apply our method to the parity task, a challenging benchmark problem, and to a chaotic system inference task. For the parity task, our symmetry-aware RC obtains zero error using an exponentially reduced artificial neurons and training data, greatly speeding up the time-to-result and outperforming hand crafted artificial neural networks (ANN). For the inference task, the performance is orders-of-magnitude better than regular RCs. We anticipate that generalizations of our procedure will have widespread applicability in information processing with ANNs.