Repository logo
 

Synthesis of neuromorphic circuits with neuromodulatory properties


Loading...
Thumbnail Image

Type

Thesis

Change log

Abstract

The field of neuromorphic engineering shows great promise in delivering novel devices inspired by biological principles that would undertake sensory and processing tasks with an unprecedented level of efficiency. In order to achieve that, engineers are required to understand and implement the many complex biological regulatory mechanisms that allow the nervous system to robustly operate and adapt over scales covering many orders of magnitude, while at the same time using unreliable and noisy components.

As a step towards that, this thesis aims at discussing and implementing the principles of neuromodulation in neuromorphic hardware, mechanisms which allow neurons to change and regulate their behaviour through the continuous control of their internal currents. We discuss how neural dynamics and its modulation can be broken down into four essential feedback loops, and we introduce a simplified model of the neural membrane respecting this fundamental structure. We present a novel methodology for controlling the neuron's behaviour through the shaping of its I-V curves in distinct timescales, thus characterising the behaviour of the neural circuit through its input-output properties. We show how modulation of the feedback loops affects the behaviour, and importantly, captures the transition between spiking and bursting oscillatory regimes, two major signalling modes of neurons. We then show how the architecture can be easily implemented using well-known neuromorphic building blocks based on subthreshold MOSFET circuits. Finally, we discuss how the excitability switch captured by the model can be exploited in simple network settings, thus opening up the possibility for future research into novel architectures where the control of cellular properties is utilised to shape the global behaviour of the network.

Description

Date

2019-10-31

Advisors

Sepulchre, Rodolphe

Keywords

Neuromorphic engineering, Neuronal behaviors, Neuromodulation, Neuronal bursting, I-V curve shaping, Subthreshold electronics

Qualification

Doctor of Philosophy (PhD)

Awarding Institution

University of Cambridge
Sponsorship
European Research Council (670645)