This website requires JavaScript.

A Note on Matrix Measure Flows, With Applications to the Contraction Analysis of Plastic Neural Networks

Leo KozachkovJean-Jacques Slotine
Dec 2022
摘要
Synapses--the connections between neurons in the brain--are constantly beingupdated. This updating is done for variety of reasons, such as helping thebrain learn new tasks or adapt to new environments. However, synapticplasticity poses a challenge for stability analyses of recurrent neuralnetworks, which typically assume fixed synapses. To help overcome thischallenge, we introduce the notion of a matrix measure flow. Given a matrixflow, a matrix measure flow captures the evolution of an associated matrixmeasure (or logarithmic norm). We show that for certain matrix flows ofinterest in computational neuroscience, the associated matrix measure flowobeys a simple inequality. This inequality can be used in turn to infer thestability and contraction properties of recurrent neural networks with plasticsynapses. We consider examples of synapses undergoing Hebbian and/orAnti-Hebbian plasticity, as well as covariance-based and gradient-based rules.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答