Source-linked AI summary
Reservoir computing with a single time-delay autonomous Boolean node
Nicholas D. Haynes, Miguel C. Soriano, David P. Rosin, Ingo Fischer, Daniel J. Gauthier
TL;DR
Reservoir computing offers an approach to pattern recognition and classification, tasks that motivate the search for biologically plausible computational paradigms. This paper tests a single autonomous Boolean logic element with time-delay feedback and finds it can classify short input patterns, with performance declining over time.
Problem
Pattern recognition and classification motivate the search for computational paradigms with biological plausibility because classical von Neumann machines cannot efficiently perform some such tasks.
Source: p. 1Method
The authors experimentally use a single autonomous XOR Boolean element with two time-delay feedback lines, characterize its effective computational dimensionality, and train linear classifiers on its output.
Source: pp. 1, 3, 4Results
Four 2-bit input patterns have error rates below 10% for approximately 70 ns, although the input word lasts only 7.5 ns.
Source: p. 4Takeaways & Limitations
The single time-delay feedback XOR reservoir produces dynamics suitable for reservoir computing and performs a pattern-recognition task.
Source: p. 4Classification error is generally higher for larger input words, and 32 possible 5-bit words cannot be accurately classified at any time.
Source: p. 4
Abstract
from arXiv · showhide
We demonstrate reservoir computing with a physical system using a single autonomous Boolean logic element with time-delay feedback. The system generates a chaotic transient with a window of consistency lasting between 30 and 300 ns, which we show is sufficient for reservoir computing. We then characterize the dependence of computational performance on system parameters to find the best operating point of the reservoir. When the best parameters are chosen, the reservoir is able to classify short input patterns with performance that decreases over time. In particular, we show that four distinct input patterns can be classified for 70 ns, even though the inputs are only provided to the reservoir for 7.5 ns.