Berlin 2015 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
BP: Fachverband Biologische Physik
BP 14: Posters: Neurophysics
BP 14.10: Poster
Monday, March 16, 2015, 17:30–19:30, Poster A
Neuronal avalanches in a self organizing recurrent neural network — •Bruno Del Papa1,2, Viola Priesemann3, and Jochen Triesch1 — 1Frankfurt Institute for Advanced Studies — 2Max Planck Institute for Brain Research — 3Max Planck Institute for Dynamics and Self-Organization
A large number of experiments have suggested that the brain operates close to criticality, but in a subcritical regime, based on power-law distributions of neuronal avalanches. Although several critical neural network models have been studied before, they typically show simplified connectivity structures and no advanced information processing or learning abilities. Here, we investigate neuronal avalanches in spontaneous activity of a self organizing recurrent neural network (SORN), which exhibits spatio-temporal pattern learning and reproduces experimentally observed fluctuations of synaptic efficacies. The network consists of excitatory and inhibitory threshold units with connection weights and firing thresholds evolving based on a combination of spike-timing dependent plasticity rules and homeostatic mechanisms. We observe power-law distributed neuronal avalanches, suggesting that the SORN self-organizes to a critical state, and find a strong dependence on the neurons' target firing rates and membrane potential noise, which indicates these are essential to maintain the critical phase. Our results show, for the first time, that signatures of criticality are present in the spontaneous activity of a self-organizing network model that has advanced learning abilities and reproduces central findings on the fluctuations of synaptic connection strengths in cortex and hippocampus.