The difference between memory and prediction in linear recurrent networks

Abstract

Recurrent networks are trained to memorize their input better, often in the hopes that such training will increase the ability of the network to predict. We show that networks designed to memorize input can be arbitrarily bad at prediction. We also find, for several types of inputs, that one-node networks optimized for prediction are nearly at upper bounds on predictive capacity given by Wiener filters, and are roughly equivalent in performance to randomly generated five-node networks. Our results suggest that maximizing memory capacity leads to very different networks than maximizing predictive capacity, and that optimizing recurrent weights can decrease reservoir size by half an order of magnitude.

References in corpus (1)

• Nearly extensive sequential memory lifetime achieved by coupled nonlinear neurons

Cited by in corpus (9)

• Scalable photonic platform for real-time quantum reservoir computing
• Exploiting Multiple Timescales in Hierarchical Echo State Networks
• Optimal short-term memory before the edge of chaos in driven random recurrent networks
• Learn to Synchronize, Synchronize to Learn
• Risk bounds for reservoir computing
• Stability Analysis of Reservoir Computers Dynamics via Lyapunov Functions
• Time cells might be optimized for predictive capacity, not redundancy reduction or memory capacity
• New analytic formulae for memory and prediction functions in reservoir computers with time delays
• Memory Uncertainty Relation and Harmonic Memory in Random Recurrent Networks