Mean Field Approach for Configuring Population Dynamics on a Biohybrid Neuromorphic System
- PDF / 1,774,359 Bytes
- 19 Pages / 595.224 x 790.955 pts Page_size
- 28 Downloads / 173 Views
Mean Field Approach for Configuring Population Dynamics on a Biohybrid Neuromorphic System 1· ¨ Johannes Partzsch1 · Christian Mayr1 · Massimiliano Giulioni2 · Marko Noack3 · Stefan Hanzsche 1 1 4 1 ¨ ¨ Stefan Scholze · Sebastian Hoppner · Paolo Del Giudice · Rene Schuffny
Received: 24 October 2019 / Revised: 6 May 2020 / Accepted: 20 May 2020 © The Author(s) 2020
Abstract Real-time coupling of cell cultures to neuromorphic circuits necessitates a neuromorphic network that replicates biological behaviour both on a per-neuron and on a population basis, with a network size comparable to the culture. We present a large neuromorphic system composed of 9 chips, with overall 2880 neurons and 144M conductance-based synapses. As they are realized in a robust switched-capacitor fashion, individual neurons and synapses can be configured to replicate with high fidelity a wide range of biologically realistic behaviour. In contrast to other exploration/heuristics-based approaches, we employ a theory-guided mesoscopic approach to configure the overall network to a range of bursting behaviours, thus replicating the statistics of our targeted in-vitro network. The mesoscopic approach has implications beyond our proposed biohybrid, as it allows a targeted exploration of the behavioural space, which is a non-trivial task especially in large, recurrent networks. Keywords Neuromorphic system · Biohybrid · Mesoscopic characterization · Switched capacitor · Mean field
1 Introduction Neuromorphic designs try to emulate the dynamic behaviour of biological neurons in CMOS circuits, with e.g. time dependent integration of synaptic inputs replicated [1]. In this they are in contrast with the new wave of circuits for deep neural network acceleration, as these only carry out a very abstracted, scalar and static numerical approximation of neurons and synapses [2]. As they provide biologically realistic behaviour, real-time neuromorphic systems allow for a direct coupling with biological Johannes Partzsch and Christian Mayr contributed equally to this work. Johannes Partzsch
[email protected] 1
Chair for Highly Parallel VLSI Systems and Neuromorphic Circuits, Department of Electrical Engineering and Information Technology, Technische Universit¨at Dresden, Dresden, Germany
2
IMASENIC Advance Imaging s.l., Barcelona, Spain
3
Ferroelectric Memory GmbH, Maria-Reiche-Str. 3, 01109 Dresden, Germany
4
Department of Technologies and Health, Istituto Superiore di Sanita, Roma, Italy
tissue [3, 4], enabling to understand, gently control and virtually extend the biological part. Seamless dynamical integration of hardware and biology makes such a hybrid system most effective, where we define seamless as that the hardware neural network operates in the same dynamical regime as its biological counterpart, and tight coupling of both generates a meaningful joint dynamics. Biohybrids can be employed to develop novel strategies for interacting with neuronal tissue, for e.g. the next generation of neuroprostheses. Biohybrids also enhance
Data Loading...
