神经形态工程学
线性
对称(几何)
机制(生物学)
计算机科学
物理
人工智能
人工神经网络
数学
量子力学
几何学
作者
Bidyabhusan Kundu,Sreetosh Goswami
摘要
From the very inception of neuromorphic computing, the quest for linear and symmetric weight updates has lingered as its most coveted yet untamed ambition. We now unveil a kinetically tuned, molecular-level mechanism that enables conductance modulation with near-ideal linearity across 16,500 analog states spanning four orders of magnitude. By orchestrating inherently nonlinear phenomena, such as nucleation, within finely controlled small-perturbation regimes, we realize what once seemed paradoxical: linearity emerging from nonlinearity. This advance offers a generalizable blueprint for instilling precise synaptic control into the very fabric of future neuromorphic materials.
科研通智能强力驱动
Strongly Powered by AbleSci AI