A 128 channel Extreme Learning Machine based Neural Decoder for Brain Machine Interfaces
Chen, Yao, Basu · cs.LG,cs.HC · 2015-09-27 · 原文
Currently, state-of-the-art motor intention decoding algorithms in brain-machine interfaces are mostly implemented on a PC and consume significant amount of power. A machine learning co-processor in 0.35um CMOS for motor intention decoding in brain-machine interfaces is presented in this paper. Using Extreme Learning Machine algorithm and low-power analog processing, it achieves an energy efficiency of 290 GMACs/W at a classification rate of 50 Hz. The learning in second stage and corresponding digitally stored coefficients are used to increase robustness of the core analog processor. The chip is verified with neural data recorded in monkey finger movements experiment, achieving a decoding accuracy of 99.3% for movement type. The same co-processor is also used to decode time of movement from asynchronous neural spikes. With time-delayed feature dimension enhancement, the classification accuracy can be increased by 5% with limited number of input channels. Further, a sparsity promoting training scheme enables reduction of number of programmable weights by ~2X.
讲义
讲义·推断 依据「原文」自动生成的结构化摘要（推断），非原文表述；以原文为准。
1. 人话版
Currently, state-of-the-art motor intention decoding algorithms in brain-machine interfaces are mostly implemented on a PC and consume significant amount of power.
A machine learning co-processor in 0.35um CMOS for motor intention decoding in brain-machine interfaces is presented in this paper.
2. 领域脉络
本文类目：cs.LG、cs.HC，属于其所在研究脉络的最新进展。
3. 机制拆解
The learning in second stage and corresponding digitally stored coefficients are used to increase robustness of the core analog processor.
4. 证据与数字
Using Extreme Learning Machine algorithm and low-power analog processing, it achieves an energy efficiency of 290 GMACs/W at a classification rate of 50 Hz.
The chip is verified with neural data recorded in monkey finger movements experiment, achieving a decoding accuracy of 99.3% for movement type.
With time-delayed feature dimension enhancement, the classification accuracy can be increased by 5% with limited number of input channels.
5. 反例与边界
摘要未声明局限与反例——这是需要警惕的信号，精读时先问边界。
6. 跨领域连接与意外收获
横跨 2 个类目（cs.LG、cs.HC），关注其在你兴趣板块间的迁移面。
7. 可复用方法
把本文机制与你手头项目对照，找一个两周内能验证的最小实验。
8. 术语表
精读时把不熟的术语记入此处，作为下次回忆的锚点。