Work

Innovations in arithmetic for neuromorphic computing

Year: 2026
Type: dissertation
Abstract: The continued growth in computational demand, driven in large part by machine learning workloads, has coincided with the slowing of performance gains from semiconductor scaling alone.Therefore, additional computing power can no longer be simply obtained by using increased transistor density or clock frequency to scale standard architecture.Instead, alternative computing paradigms are required.Neuromorphic computing has emerged as one such approach, offering architectures designed to improve computational efficiency by mimicking aspects of biological neural networks such as massive parallelism, locality of memory and computation, and sparse activity.Spiking Neural Networks (SNNs) are the primary computational models executed on neuromorphic processors, processing information through temporal dynamics and sparse events rather than dense numerical operations.However, the numerical representation used to implement these dynamics remains an open design question.This thesis investigates the role of arithmetic choice in digital neuromorphic computing, with a particular focus on posit arithmetic as an alternative to conventional fixed-point and IEEE-754 floating-point formats.Rather than treating reduced precision as an optimisation problem to be mitigated against, this work adopts an arithmetic-centric perspective, evaluating whether alternative number systems can intrinsically offer improved numerical behaviour at reduced bit-widths.To this end, rather than the more common mixed-precision approach, all experiments use reduced-precision emulation throughout to simulate a neuromorphic system operating natively with the target format, providing a conservative assessment of arithmetic suitability for future neuromorphic hardware.The thesis makes three primary contributions.First, it presents a systematic study of reduced-precision arithmetic for simulating neural dynamics, using the Izhikevich neuron model across its full range of canonical firing patterns.The results demonstrate that 16bit posit arithmetic can match or outperform IEEE-754 floating-point equivalents of equal bit-width, and in some cases 16-bit posit can match 64-bit floating-point accuracy, if combined with a simple rescaling strategy.Second, the work provides the first comprehensive evaluation of 8-bit posit arithmetic for offline SNN training using gradient-based methods.Across both frame-based and event-based datasets, 8-bit posits are shown to enable effective training without the need for mixed-precision schemes, loss scaling, or other mitigation techniques commonly required by low-precision floating-point formats.Finally, the thesis presents the first detailed analysis of arithmetic effects on online learning in SNNs using the e-prop algorithm, demonstrating that 16-bit posits enable stable training and outperform higher-precision floating-point formats, while fixed-point representations require approximately 64-bit precision to attain similar performance.Collectively, these results show that posit arithmetic can match the numerical and system level performance of larger floating point formats while using smaller bit widths and requiring fewer, if any, mitigation strategies.The designers of future digital neuromorphic architectures should therefore seriously consider including posit rather than fixed or floating point arithmetic units. ThanksI would not have made it this far without the help and support of quite a few special people.In a very particular order... My partner Rachel, who agreed to take on the mantle of family breadwinner, allowing me to take a colossal pay cut and become a student again, to study, retrain and start my (probably) final career (no promises).Without her sacrifices over the last 4 years, the whole endeavour would have been a none starter.Sorry for being a nightmare to varying degrees over that time.I don't know how you put up with me.I love you loads Rachel, always will, thank you.A huge thanks to Tatiana who has joined me on this roller-coaster right from the start.For overcoming your initial doubts and taking me on.Encouraging me when I lost faith, and pestering me with just the right amount of tenacity throughout.Generous and patient, I couldn't have done it without you.Tatiana -thank you.Another huge thank you to Jamie.Your peerless knowledge of the inner workings of a neuromorphic processor has been, and continues to be invaluable.Full of encouragement and always ready to help, with endless patience wading through pages of my unpossibly bad English.Jamie -thank you.Thank you to everyone at Brunel, past and present members of the AI research group.Accepting me in and treating me as an equal making me feel welcome!To my industrial sponsors: Flemming and Sundance for introducing me to Tatiana, providing funding for the first two years and getting the ball rolling.Even greater thanks to Ray and Master Corp. who stepped in when all looked lost, to fund the final year with no immediate benefit for you.I will never forget.Of course my mum and dad, the only people outside academia who will listen to me bang on about computer arithmetic without falling asleep, or understanding a word.Your unwavering belief in me, has kept me going!Lastly, my children Isaac & Flo.Isaac, love you loads son.Embrace your stretch zone and you'll be taller than me in every way.You are amazing.And Flo, for being a full-force, unicorn festooned, rainbow of boundless energy and happiness.Daddy has finished his work now Flo, thank you so much for being patient with me, sorry it took ages.Come on, we can go to the park now Nomenclature
Language: English
Open Access status: green
OpenAlex ID: W7204375939
Cites: 263
Cited by:
Related to: 10