Item

Abstract

Theoretical and experimental evidence brought forward a hypothesis that the brain operates close to a critical state. Numerous studies investigated neural models that can attain various distances to criticality depending on a control parameter and quantified information processing capabilities as a function of closeness to criticality. However, quantifying these capabilities in a general sense is not sufficient to assure usefulness of criticality for the brain. Therefore, we introduce a complementary approach. We study a network that is optimized for a task relevant for the brain. Then, we investigate whether we observe the scale-free neuronal avalanches exclusively in the optimized network. More specifically, we used a network of leaky integrate-and-fire neurons with parameters optimized for efficient coding. Previously, it was shown that performance of such networks varies non-monotonically with the noise amplitude. We discovered, that only in the network with optimal noise level the avalanche size distribution follows a power-law and with too low or too high noise, the network appears either super-critical or sub-critical, respectively. We demonstrate that scale-free distribution of neuronal avalanches might be a consequence of optimal efficient coding in spiking neural networks. This result has important implications, as it shows how two influential, and previously disparate fields - efficient coding, and criticality - might be intimately related.