| 研究実績の概要 |
In the last academic year, we have finalized the project by publishing the final result. This project is motivated by evidence indicating that hippocampal adult neurogenesis is vital for distinguishing highly similar memories. During observed neural process, synaptic competition along with long-term potentiation adjusts the weights of synaptic connections across neurons in a nonlocal manner, offering a distinct form of unsupervised learning compared to Hebb’s local plasticity rule.
However, the mechanism by which synaptic competition facilitates the separation of similar memories remains largely unexplored. In this project, we aim to connect synaptic competition with pattern separation. Adult-born neurons integrate into the existing neuronal network by competing with mature neurons for synaptic connections from the entorhinal cortex. Our findings reveal that synaptic competition and neuronal maturation have unique roles in differentiating overlapping memory patterns. Moreover, we show that a feedforward neural network trained with a competition-based learning rule can surpass a multilayer perceptron trained with the backpropagation algorithm, especially when limited training samples are available. These results illuminate the functional significance and potential applications of synaptic competition in neural computation. A manuscript reporting these results is published in PNAS Nexus.
|
