Learning in Biologically Plausible Neural Networks

Yunlong Xu

Abstract: Biologically Plausible Neural Networks (BPNNs) have attracted significant attention in recent years, mainly due to their ability to bridge the gap between artificial neural networks and the biological processes that underlie them. In this paper, I present an exhaustive literature review of learning algorithms for three specific types of BPNNs: 1) Constrained Deep Neural Networks (CDNNs), 2) Spiking Neural Networks (SNNs), and 3) Reduced Spiking Neural Networks (Rate Models, RSNNs). Through case studies, I demonstrate the implementation and training of CDNNs and introduce a novel learning method for RSNNs, which we also implement. Furthermore, I propose an innovative approach to compare Spiking Neural Networks and Constrained Deep Neural Networks. As future work, I plan to expand my investigation of learning algorithms for SNNs, an endeavor that will further enhance our understanding of biologically inspired neural network models.