Simulation, visualization and analysis tools for pattern recognition assessment with spiking neuronal networks
Computational modeling is becoming a widely used methodology in modern neuroscience. However, as the complexity of the phenomena under study increases, the analysis of the results emerging from the simulations concomitantly becomes more challenging. In particular, the configuration and validation of brain circuits involving learning often require the processing of large amounts of action potentials and their comparison to the stimulation being presented to the input of the system. In this study we present a systematic work-flow for the configuration of spiking-neuronal-network-based learning systems including evolutionary algorithms for information transmission optimization, advanced visualization tools for the validation of the best suitable configuration and customized scripts for final quantitative evaluation of the learning capabilities. By integrating both grouped action potential information and stimulation-related events, the proposed visualization framework provides qualitatively assessment of the evolution of the learning process in the simulation under study. The proposed work-flow has been used to study how receptive fields emerge in a network of inhibitory interneurons with excitatory and inhibitory spike-timing dependent plasticity when it is exposed to repetitive and partially overlapped stimulation patterns. According to our results, the output population reliably detected the presence of the stimulation patterns, even when the fan-in ratio of the interneurons was considerably restricted.