The spiking neural network that was developed as part of this thesis has 17,544 neurons and 698,625 connections, and it controls the eye movements of the SIMNOS virtual robot. A detailed description of this network is given in Chapter 5 and it was simulated using the SpikeStream simulator. A diagram of the network is given in Figure 1.
Figure 1. Neural network
During the training phase, the network spontaneously generates eye movements to different parts of its visual field and learns the association between an eye movement and a visual stimulus. After training, Motor Cortex moves SIMNOS's eye around at random until a blue object appears in its visual field. This switches the network into its offline 'imagination' mode, in which it generates motor patterns and 'imagines' the red or blue visual input that is associated with these potential eye movements. This process continues until it 'imagines' a red visual stimulus that positively stimulates Emotion. This removes the inhibition, and SIMNOS's eye is moved to look at the red stimulus.
Videos of the network in operation are available here.
SpikeStream file of the trained network: TrainedNeuralNetwork.sql.tar.gz.
Analysis Run 1 and Noise Run 1
SpikeStream archive of the network in operation: AnalysisRun1_NoiseRun1_NeuralArchive.sql.tar.gz. This archive contains Analysis Run 1 and Noise Run 1, which are described in Chapter 7, Section 7.9.2. A video of Analysis Run 1 is available and an XML description of the predicted phenomenology of the network during Analysis Run 1 and Noise Run 1 can be accessed here.