Skip to main content

PhD project title

Biologically-plausible Neural Processors

Outline description, including interdisciplinary, intersectoral and international dimensions

Emulating the human brain is one of the grand challenges of modern-day science. It is seen as critically important for neuroscientists studying how learning and cognitive behaviours and disease arise in humans. Similarly, in applications such as advanced manufacturing, it underpins attempts to develop systems which respond to electronic stimuli and intelligently enact electronic or robotic actions accordingly.

In both these cases, as biological brains cannot be created, computer-based emulations are used. However, while the human brain houses approximately 86 billion () neurons of many different types of neuron alongside 1 quadrillion (1) synapses, current emulations support only a tiny fraction of this. State-of-the-art applications, such as SPAUN1, house of the order of millions of neurons – a tiny proportion of the human brain.

New computer architectures and engineering approaches are emerging to address this challenge of scale and diversity. Digital neuromorphic computers, such as SpiNNaker3 or BrainFrame3 support different neuron behaviours but are extremely inefficient and limited by their physical size and power consumption. Analogue alternatives such as Braindrop4 or TrueNorth5 are comparatively high efficiency, but can only realise a single neuron type each.

This project aims to enable a breakthrough solution which enables both the flexibility of digital neuromorphic devices with the efficiency of analogue. It will:

  1. Build on leading expertise in memristor technologies in the Centre for Nanostructured Media in School of Maths & Physics (MAP) to create analogue neuronal circuits which are highly efficient but also support different neuron types.
  2. Build on existing expertise in ECIT’s Centre of Data Science and Scalable Computing (DSSC) in programmable hardware to integrate these neurons into a programmable processor architecture.
  3. Integrate 1 & 2 with the Neural Engineering Framework (NEF)6 to allow it to realise custom spiking neural network workloads.
  1. Eliasmith et al., “A Large-Scale Model of the Functioning Brain”, Science, 338 (2012), pp. 1202-1205.
  2. B. Furber, F. Galluppi, S. Temple and L. A. Plana, "The SpiNNaker Project," Proceedings of the IEEE, 102 (2014) pp. 652-665.
  3. Smaragdos et al., “BrainFrame: a node-level heterogeneous accelerator platform for neuron simulations”, Journal of Neural Engineering, 14 (2017).
  4. Neckar et al., "Braindrop: A Mixed-Signal Neuromorphic Architecture With a Dynamical Systems-Based Programming Model," Proceedings of the IEEE, 107 (2019), pp. 144-164.
  5. Akopyan et al., "TrueNorth: Design and Tool Flow of a 65 mW 1 Million Neuron Programmable Neurosynaptic Chip," IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 34 (2015), pp. 1537-1557.
  6. Bekolay et al., “Nengo: a Python Tool of Building Large-scale Functional Brain Models”, Frontiers in Neuroinformatics, 7 (2014), pp. 1662 – 5196.

Key words/descriptors


Artificial Intelligence, Spiking Neural Networks, Data Science, Neuromorphic, Memristor.

Fit to CITI-GENS theme(s)

  • Information Technology
  • Advanced Manufacturing

Supervisor Information



First Supervisor: Dr. John McAllister                                                                 School: EEECS

Second Supervisor:   Prof. Marty Gregg                                                           School: MAP

Third Supervisor:  Bernd Gotsmann                                                                 Company: IBM Zurich

Name of non-HEI partner(s)

We expect input from IBM Zurich, as Prof Gregg has ongoing activities on memristors with them as official non-HEI partners.

Contribution of non-HEI partner(s) to the project:



We expect context and guidance for the neuromorphic research performed at QUB against activity in IBM. We expect that the fellow will be able to spend some time at IBM Zurich, as this is the agreement for an ESR who will join Prof Gregg’s group over the summertime in 2020, working on the memristor theme. Extending the opportunity for a stay in the IBM labs to this CITI-GENS fellow should be a formality.