Neural network modelling of the influence of channelopathies on reflex visual attention

• Main Author: Gravier, Alexandre
• Other Authors: Włodzisław Duch
• Format: Theses and Dissertations
• Language: English
• Published: 2015
• Subjects: DRNTU::Science::Biological sciences::Human anatomy and physiology::Neurobiology; DRNTU::Engineering::Computer science and engineering::Computer applications::Life and medical sciences; DRNTU::Engineering::Mathematics and analysis::Simulations
• Online Access: https://hdl.handle.net/10356/65485
• Institution: Nanyang Technological University

Research in computational neuroscience has the potential to bridge the gap between the neuroscientific descriptions of neural structures and dynamics, and the psychological models of behaviour and mental states. In particular, simulated neural network models with output or dynamics that are interpretable in terms of behaviour can help to understand neurodynamics and assess the plausibility of vertical theories of cognition. This thesis presents two contributions to the domain: EVAC, a model of emergent visual attention in presence of calcium channelopathy, and PyCogMo, a model learning framework. By modelling channelopathy, EVAC constitutes an effort towards identifying the possible causes of autism. The network structure embodies the dual pathways model of cortical processing of visual input, with reflex attention as an emergent property of neural interactions. EVAC extends a model developed by O'Reilly and Munakata [1] by introducing attention shift in a larger-scale network and applying a phenomenological model of channelopathy. The simulation of EVAC results in testable behavioural predictions on reflex attention shift tasks inspired by Posner [2]. In presence of a distractor, the channelopathic network's rate of failure to shift attention is lower than the control network's, but overall, the control network exhibits a lower classification error rate. The simulation results also show differences in task-relative reaction times between control and channelopathic networks. The attention shift timings inferred from the model are consistent with studies of attention shift in autistic children. Often, generative methods, like the simulation of the model of EVAC, require to train the synaptic weights using specialised neural networks learning algorithms. In the EVAC model, the weights are learnt with the Leabra algorithm \cite{cecn}. Emergent is the only neural simulator implementing the Leabra algorithm. Consequently, the EVAC network could not be developed in a simulator-independent manner that would allow it to be run on a variety of implementations to improve the reliability of the results. While PyNN [3] is a software framework that offers the possibility to specify and run a model on several otherwise incompatible simulators, training those networks remains a simulator-dependent task, for which PyNN is not suited. Hence, PyCogMo, a novel learning framework based on PyNN, was developed to address this limitation. It enables modellers to implement and use learning algorithms with several simulators. PyCogMo constitutes a step towards improving the reproducibility of computational experiments across simulators, hence a valuable methodological support for computational neuroscientists.