Use this URL to cite or link to this record in EThOS:
Title: A reconfigurable neuroprocessor platform for biophysically accurate neural networks
Author: Moctezuma Eugenio, Juan Carlos
ISNI:       0000 0004 5921 7829
Awarding Body: University of Bristol
Current Institution: University of Bristol
Date of Award: 2015
Availability of Full Text:
Access from EThOS:
In computational neuroscience, the modelling of neurons at biophysically accurate level provides an important tool to understand how the brain works. In contrast with reduced neuron models, which try to save computational resources and simplify neuron behaviour. Biophysical accurate systems based on conductance-based models can capture the dynamics of real neurons in more detail and provide biological meaningful information. Most of the work done in this area focuses on developing platforms that emulate partial or complete brain's behaviour, requiring the simulation of large neural networks using simplified models. Research in single neurons dynamics and small neuro-systems at a biophysical level of abstraction has been neglected and not sufficiently explored; mainly due to the heavy computational cost that biophysical accurate modelling entails. The study of such size-constrain systems is crucial to understand the nervous system in many biological organisms presented in nature and the realization of simulation platforms to support such systems is needed. On this thesis, it is presented a reconfigurable neuro-simulator platform specially designed to emulate biophysical accurate and biological compatible neural networks. The platform is based on FPGA technology that allows the flexibility to create custom neuroprocessors architectures to solve conductance-based models at real time with floating point accuracy. Through a series of experiments the dynamics of neuroprocessors are evaluated and compared with real neuron responses. The problem of interconnecting neurons with individual synapses was tackled with a novel synaptic architecture where all incoming synapses are merged efficiently in one single accumulative process without losing biological information. These results suggest the suitability of conductance-based models and FPGA platforms to simulate living organisms' behaviour in a biological compatible context.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available