忆阻器类脑芯片与人工智能( 七 ) 来源：文章转载自期刊《微纳

3.类脑芯片与人工智能系统硬件化
现阶段的人工智能大多是基于软件实现的，其较高的功耗需求成为智能终端化发展道路上的障碍。下面以实现人脸识别为例展示基于类脑芯片的人工智能芯片在功耗和速率方面的优势。基于软件的人脸识别的流程是：首先将所有目标人脸图像数据化存储在存储器中，同时存储其对应的身份信息；进行人脸图像识别时，将图像信息数据化后和存储器中所存储的所有图像信息进行对比，找到匹配度最高的图像；输出图像信息对应的身份信息。由此可知，每次进行图像识别时，拟识别图像要和所有存储图像数据进行对比，随着存储图像数据的增加，功耗的增加和速率的下降是无法忽视的。未来基于类脑芯片的人脸图像识别，其识别流程如下：首先将人脸图像数据化后输入到类脑芯片中，通过修改类脑芯片中类突触器件的参数，使在输入不同图像数据时，输出端口有不同的输出值，将不同输出值和输入图像的身份信息对应并存储，实现对类脑芯片的训练；在识别图像时，将图形数据化后输入到类脑芯片中，类脑芯片直接输出图像对应的身份信息。由此可知，只要类脑芯片训练完成，其能够在及低功耗下快速给出拟识别图像的身份信息，不需要循环对比计算。综上，基于忆阻器等类突触器件的类脑芯片能够促进人工智能系统硬件化、终端化的早日实现。
文献引用：
陈子龙，程传同，董毅博，等. 忆阻器类脑芯片与人工智能[J]. 微纳电子与智能制造, 2019, 1(4): 58-70.
CHEN Zilong, CHENG Chuantong , DONG Yibo, et al. Brain-like chips and artificial intelligence based on memristors[J]. Micro/nano Electronics and Intelligent Manufacturing, 2019, 1(4): 58-70.
《微纳电子与智能制造》刊号：CN10-1594/TN
主管单位：北京电子控股有限责任公司
主办单位：北京市电子科技科技情报研究所
北京方略信息科技有限公司
参考文献:
[1] WONG HSP, LEE HY, YU S, et al. Metal-oxide RRAM [J]. Proceedings of the IEEE, 2012, 100(6): 1951-1970.
[2] CHENG C, ZHU C, HUANG B, et al. Processing halide perovskite materials with semiconductor technology[J]. Advanced Materials Technologies, 2019, 4(7): 1800729.
[3] HIROSE Y, HIROSE H. Polarity- dependent memory switching and behavior of Ag dendrite in Ag- photodoped smorphous As2S3 films[J]. Journal of Physics D, 1976, 47: 2767-2772.
[4] TERABE K, HASEGAWA T, NAKAYAMA T, et al. Quantized conductance atomic switch[J]. Nature, 2005, 433(7021): 47-50.
[5] KOZICKI M N, Park M, MITKOVA. Nanoscale memory elements based on Solid- State electrolytes[J]. IEEE Transactions on Nanotechnology, 2005, 4(3): 331-338.
[6] HENISCH HK, SMITH W R. Switching in polystyrene films[J]. Applied Physics Letters, 1974, 24: 589-591.
[7] SEGUI Y, AI B, CARCHANO H. Switching in polystyrene films: transition from on to off state[J]. Journal of Applied Physics, 2008, 47(1): 140-143.
[8] VERBAKEL F, MESKERS SCJ, JANSSEN R. Electronic memory effects in a Sexithiophene-Poly(ethylene oxide) block copolymer doped with NaCl. combined diode and resistive switching behavior[J]. Chemistry of Materials, 2006, 18(11): 2707-2712.
[9] STRUKOV D B, SNIDER G S, Stewart D R, et. al. The missing memristor found[J]. Nature ， 2008 ， 453: 80-83.
[10] CHANG T, JO SH, KIM KH, et al. Synaptic behaviors and modeling of a metal oxide memristive device[J]. Applied Physics A, 2010, 102(4): 857-863.
[11] WANG Z, JOSHI S , SAVEL’EV, SERGEY, et al. Fully memristive neural networks for pattern classification with unsupervised learning[J]. Nature Electronics, 2018, 1(2): 137-145.
[12] SIEMON A, MENZEL S, WASER R, et al. A complementary resistive switch- based crossbar array adder[J]. IEEE Journal on Emerging and Selected Topics in Circuits and Systems, 2015, 5(1): 64-74.
[13] MERRIKH B F, PREZIOSO M, CHAKRABARTI B, et al. Implementation of multilayer perceptron network with highly uniform passive memristive crossbar circuits [J]. Nature Communications, 2018, 9(1): 2331.
[14] YAO P, WU H, GAO B, et al. Face classification using electronic synapses[J]. Nature Communications, 2017 (8): 15199.
[15] YANG R, HUANG H M, HONG Q H, et al. Synaptic suppression Triplet- STDP learning rule realized in Second- Order memristors[J]. Advanced Functional Materials, 2017, 28(5): 1704455.
[16] WAN C J, ZHU L Q, ZHOU J M, et al. Inorganic proton conducting electrolyte coupled oxide- based dendritic transistors for synaptic electronics[J]. Nanoscale, 2014, 6 (9): 4491-4497.
[17] XU X, LUO Q, GONG T, et al. Fully CMOS compatible 3D vertical RRAM with self- aligned self- selective cell enabling sub-5nm scaling[C]// IEEE Symposium on VL Technology. IEEE, 2016.
[18] WANG M, CAI S, PAN C, et al. Robust memristors based on layered two- dimensional materials[J]. Nature Electronics, 2018, 1: 130-136.