Doborjeh, MaryamLind, Andrew2024-10-302024-10-302024http://hdl.handle.net/10292/18206This thesis presents a novel approach to computational modelling that seeks to enhance Spiking Neural Network (SNN) models by simultaneously incorporating genetic information and Neuroreceptor-Dependent Plasticity (NRDP) mechanisms. Motivated by the need to develop biologically inspired computational models that accurately capture the interplay between genetic factors and neural activity, the goal is to establish a more holistic computational paradigm. A biologically enhanced SNN model that integrates genetic information into the underlying Spiking Neuron model, and implements an NRDP unsupervised learning rule based on a computational model of NRDP dynamics is introduced. Integration of the genetic information is facilitated via a novel and simplified Gene Regulatory Network (GRN) model, and a generalised version of the Leaky Integrate-and-Fire (LIF) Spiking Neuron model. Based on experimental validation of this biologically enhanced SNN model using Electroencephalogram (EEG) data from Alzheimer’s Disease patients and controls, the findings reveal a significant improvement in classification accuracy over baselined methods. These findings also suggest enhanced synaptic connectivity, particularly within the hippocampal region, resulting in an improvement in classification accuracy of 14.7%. Improvements in the network's interpretative capabilities are also demonstrated, with an uplift of 6.8% in specificity, enabling the model to learn context-dependent associations and make nuanced interpretations. Finally, enhanced contextual recall is observed, increasing sensitivity by 14.8%, allowing the model to utilise past experiences for current interpretations, thus making its decisions more explainable and leading to more favourable knowledge discovery. These results underscore the potential of genetically informed SNNs to more accurately reflect biological processes, thereby advancing the field of biologically inspired computing, and taking important steps towards biologically plausible Artificial Neural Networks.enThesisOpenAccess