Abstract

FZJ-2020-00876

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png Reservoir Transfer on Analog Neuromorphic Hardware

He, X.FZJ* ; Liu, T. ; Hadaeghi, F. ; Jaeger, H.

2019

This record in other databases:  

Please use a persistent id in citations: doi:10.1109/NER.2019.8716891

Abstract: Analog, unclocked, spiking neuromorphic microchips open new perspectives for implantable or wearable biosensors and biocontrollers, due to their low energy consumption and heat dissipation. However, the challenges from a computational point of view are formidable. Here we outline our solutions to realize the reservoir computing paradigm on such hardware and address the combined problems of low bit resolution, device mismatch, approximate neuron models, and timescale mismatch. The main contribution is a computational scheme, called Reservoir Transfer, which enables us to transfer the dynamical properties of a well-performing neural network which has been optimized on a digital computer, onto neuromorphic hardware that displays the abovementioned problematic properties. Here we present a case study of implementing an ECG heartbeat abnormality detector to showcase the proposed method.

Note: Tianlin Liu was supported by the FZJ through the project SMARTSTART Computational Neuroscience, DB001423.

---

Contributing Institute(s):

1. Computational and Systems Neuroscience (INM-6)
2. Jara-Institut Brain structure-function relationships (INM-10)
3. Theoretical Neuroscience (IAS-6)

Research Program(s):

1. 574 - Theory, modelling and simulation (POF3-574) (POF3-574)
2. Smartstart - SMARTSTART Training Program in Computational Neuroscience (90251) (90251)

Appears in the scientific report 2019

---