Guided Self-Organization of Input-Driven Recurrent Neural Networks
This work addresses the optimization of reservoir computing systems for researchers in machine learning and computational neuroscience, but it is incremental as it builds on existing review and conceptual frameworks.
The paper reviews methods for understanding input-driven recurrent neural networks and reservoir computing, and demonstrates how guided self-organization can enhance reservoir performance, though specific numerical results are not provided.
We review attempts that have been made towards understanding the computational properties and mechanisms of input-driven dynamical systems like RNNs, and reservoir computing networks in particular. We provide details on methods that have been developed to give quantitative answers to the questions above. Following this, we show how self-organization may be used to improve reservoirs for better performance, in some cases guided by the measures presented before. We also present a possible way to quantify task performance using an information-theoretic approach, and finally discuss promising future directions aimed at a better understanding of how these systems perform their computations and how to best guide self-organized processes for their optimization.