Bio-Inspired Stochastic Computing Using Binary CBRAM Synapses
Bio-Inspired Stochastic Computing Using Binary CBRAM Synapses
复制标题
DOI：
10.1109/ted.2013.2263000
复制
发表时间：
2013-07-01
影响因子：
3.1
通讯作者：
DeSalvo, Barbara
中科院分区：
文献类型：
--
作者：
Suri, Manan;Querlioz, Damien;DeSalvo, Barbara
在本文中，我们提出了一种替代方法的神经形态系统的基础上，多级电阻记忆突触和确定性学习规则。我们展示了一种原始的方法，使用导电桥RAM（CBRAM）设备，易于编程和低功耗，随机学习规则的二进制突触。提出了一种新的电路结构、编程策略和概率尖峰定时相关可塑性（STDP）学习规则，用于两种不同的带选择器（1 T-1 R）和不带选择器（1 R）的CBRAM配置。我们展示了两种方法（内在和外在）实现概率STDP规则。通过两个示例应用说明了具有二进制突触的完全无监督学习：1）实时听觉模式提取（灵感来自64通道硅耳蜗模拟器）;以及2）视觉模式提取（灵感来自视觉皮层内部的处理）。高精度（音频模式灵敏度> 2，视频检测率> 95%）和低突触功耗（音频0.55 μ W，视频74.2 μ W）。分析了突触参数变化对系统性能的影响和鲁棒性。
In this paper, we present an alternative approach to neuromorphic systems based on multilevel resistive memory synapses and deterministic learning rules. We demonstrate an original methodology to use conductive-bridge RAM (CBRAM) devices as, easy to program and low-power, binary synapses with stochastic learning rules. New circuit architecture, programming strategy, and probabilistic spike-timing dependent plasticity (STDP) learning rule for two different CBRAM configurations with-selector (1T-1R) and without-selector (1R) are proposed. We show two methods (intrinsic and extrinsic) for implementing probabilistic STDP rules. Fully unsupervised learning with binary synapses is illustrated through two example applications: 1) real-time auditory pattern extraction (inspired from a 64-channel silicon cochlea emulator); and 2) visual pattern extraction (inspired from the processing inside visual cortex). High accuracy (audio pattern sensitivity > 2, video detection rate > 95%) and low synaptic-power dissipation (audio 0.55 mu W, video 74.2 mu W) are shown. The robustness and impact of synaptic parameter variability on system performance are also analyzed.