UCL Department of Electronic and Electrical Engineering


EEE academic funded for an EPSRC #NewHorizons project

6 October 2022

UCL Electronic and Electrical Engineering’s (EEE) Dr Adnan Mehonic will explore high-risk speculative research ideas as part of a £15million initiative.

Dr Adnan Mehonic receives EPSRC funding

Funded by the Engineering and Physical Sciences Research Council (EPSRC) this year’s #NewHorizons initiative spans across 77 adventurous projects – with Dr Mehonic’s project entitled ‘Memory Impedance for Efficient Complex-valued Neural Networks’, focusing on developing a new class of nanoelectronic devices that can lead to highly energy efficient ways of computing and processing data at the edge.

On being granted funding for the project, Adnan stated:

In this innovative project, we will be developing fundamentally novel electronic nanoscale components and systems with a unique capability of directly processing complex signals. The research sits at the interface between materials science, microelectronics, energy-efficient AI and signal processing. I hope it will open new research directions toward timely, energy-efficient, highly functional AI and signal processing applications. I am very grateful to the EPSRC for supporting the research, the project co-I Prof Alex Shluger, and the project partner Prof Judith Driscoll.

As part of the project, Dr Mehonic is currently recruiting for a Research Fellow in Memristive Technology. The aim will be to develop novel analogue memimpedance devices by understanding the underpinning physics and exploring their use in complex-valued neural networks. 
Further information and to apply.

Project abstract

AI, particularly in the form of artificial neural networks (ANNs), has become indispensable in a wide range of rapidly growing data-centric technologies. However, AI has a hardware problem [Nat Electron 1, 205 - 205 (2018)] because current computing systems consume far too much energy. This is not sustainable and is rapidly becoming a critical societal challenge. The soaring demand for computing power vastly outpaces improvements made through Moore’s scaling or innovative architectural solutions - the computing demands of AI applications now double every 2 months. As a direct consequence, the real cost of training state-of-the-art AI models increased exponentially, from a few $ in 2012 to ~$10m in 2020 [Big Ideas ‘21, Ark Invest]. The challenge is even more pronounced where energy resources are limited (e.g. edge computing in the IoT era). A pressing need to develop novel technologies to address this issue at the fundamental level and build efficient AI systems has recently become acute. This risky and adventurous project will solve the problem by developing a novel class of electronic devices called memimpedors and exploring their use in complex-valued neural networks. Memimpedors are generalised memristors that enable seamless and extremely energy-efficient direct processing at the edge.

Additional information
EPSRC press release