小柯机器人

美国哥伦比亚大学Nathaniel B. Sawtell团队近日取得一项新成果。他们揭示在电鱼的多层网络中持续学习行为。2019年11月14日出版的《细胞》发表了这项成果。

研究人员提供了一个关于mormyrid鱼的电感应叶（ELL）中多个处理层的学习信息，并报告了它如何解决机器学习中众所周知的问题。由于ELL不断进行操作和学习，因此它必须协调学习和信令功能，而无需切换其操作模式。研究人员发现，这是通过在中间层神经元内进行功能区分开来实现的，其中驱动学习的输入差异地影响树突状和轴突突峰。研究人员还发现，基于学习而非感官反应选择性的连通性可确保突触到中间层神经元上的可塑性与输出神经元的需求相匹配。这些发现的机制与在小脑、海马和大脑皮层以及人工系统中的学习有关。

据了解，在人工神经网络中，跨多层分布式学习已被证明非常强大。但是，关于如何在大脑中实现多层学习知之甚少。

附：英文原文

Title: Continual Learning in a Multi-Layer Network of an Electric Fish

Author: Salomon Z. Muller, Abigail N. Zadina, L.F. Abbott, Nathaniel B. Sawtell

Issue&Volume: November 14, 2019

Abstract: Distributing learning across multiple layers has proven extremely powerful in artificialneural networks. However, little is known about how multi-layer learning is implementedin the brain. Here, we provide an account of learning across multiple processing layersin the electrosensory lobe (ELL) of mormyrid fish and report how it solves problemswell known from machine learning. Because the ELL operates and learns continuously,it must reconcile learning and signaling functions without switching its mode of operation.We show that this is accomplished through a functional compartmentalization withinintermediate layer neurons in which inputs driving learning differentially affectdendritic and axonal spikes. We also find that connectivity based on learning ratherthan sensory response selectivity assures that plasticity at synapses onto intermediate-layerneurons is matched to the requirements of output neurons. The mechanisms we uncoverhave relevance to learning in the cerebellum, hippocampus, and cerebral cortex, aswell as in artificial systems.

DOI: 10.1016/j.cell.2019.10.020

Source: https://www.cell.com/cell/fulltext/S0092-8674(19)31170-5