Making Sense of Neural Networks in the Light of Evolutionary Optimization

Abstract

AbstractTo what extent can evolution be considered as the sole first principle that explains all properties of nervous systems? This paper proposes an innovative, mathematically rigorous perspective on understanding nervous systems from an evolutionary perspective, leveraging methods of nonequilibrium statistical physics. This approach allows for modeling an exhaustive evolutionary optimization of nervous systems irrespective of any specific molecular and cellular mechanisms, approximate neuronal models or evolutionary history events. This novel method may shed light on key properties of biological neural networks and may also have potential relevance for understanding artificial neural networks.