KNOWM Memristors in a Bridge Synapse delay-based Reservoir Computing system for detection of epileptic seizures

TL;DR

This paper demonstrates a neuromorphic reservoir computing system using KNOWM memristors for epileptic seizure detection, showcasing their potential in reliable, low-power hardware implementations for complex time series classification.

Contribution

It introduces a single-node Echo State Machine based on bridge memristor synapses for seizure detection, highlighting the practical application of KNOWM memristors in neuromorphic computing.

Findings

Successful seizure detection using memristor-based reservoir computing

Analysis of memristor switching and internal dynamics

Effective classification of complex time series data

Abstract

Nanodevices that show the potential for non-linear transformation of electrical signals and various forms of memory can be successfully used in new computational paradigms, such as neuromorphic or reservoir computing (RC). Dedicated hardware implementations based on functional neuromorphic structures significantly reduce energy consumption and/or increase computational capabilities of a given artificial neural network system. Concepts of RC, which as a flexible computational paradigm can be highly inclusive, are often used as a model to describe computations performed in materia. With mostly fixed internal structure, solid-state devices, especially memristors, are studied as computational substrates in various RC systems. In this work, we present single-node Echo State Machine (SNESM) RC system based on bridge synapse as a computational substrate (consisting of 4 memristors and a differential amplifier) used for epileptic seizure detection. KNOWM memristors were posed as ideal candidates because of their easy prototyping and reliability of operation. In this account, we present an application of commercially available KNOWM memristors in various neuromorphic applications, from simple analysis of switching and internal dynamics (elucidated form noise spectroscopy and total harmonic distortion analysis) to the classification and recognition of complex time series: epilepsy seizure recognition using a wrist-worn triaxial accelerometer.