000136161 001__ 136161
000136161 005__ 20251113150204.0
000136161 0247_ $$2doi$$a10.1109/TETC.2024.3406628
000136161 0248_ $$2sideral$$a138866
000136161 037__ $$aART-2024-138866
000136161 041__ $$aeng
000136161 100__ $$aZahedi, Mahdi
000136161 245__ $$aBCIM: Efficient Implementation of Binary Neural Network Based on Computation in Memory
000136161 260__ $$c2024
000136161 5060_ $$aAccess copy available to the general public$$fUnrestricted
000136161 5203_ $$aApplications of Binary Neural Networks (BNNs) are promising for embedded systems with hard constraints on energy and computing power. Contrary to conventional neural networks using floating-point datatypes, BNNs use binarized weights and activations to reduce memory and computation requirements. Memristors, emerging non-volatile memory devices, show great potential as a target implementation platform for BNNs by integrating storage and compute units. However, the efficiency of this hardware highly depends on how the network is mapped and executed on these devices. In this paper, we propose an efficient implementation of XNOR-based BNN to maximize parallelization. In this implementation, costly analog-to-digital converters are replaced with sense amplifiers with custom reference(s) to generate activation values. Besides, a novel mapping is introduced to minimize the overhead of data communication between convolution layers mapped to different memristor crossbars. This comes with extensive analytical and simulation-based analysis to evaluate the implication of different design choices considering the accuracy of the network. The results show that our approach achieves up to 5 × energy-saving and 100 × improvement in latency compared to baselines.
000136161 540__ $$9info:eu-repo/semantics/openAccess$$aby-nc-nd$$uhttps://creativecommons.org/licenses/by-nc-nd/4.0/deed.es
000136161 590__ $$a5.4$$b2024
000136161 592__ $$a1.284$$b2024
000136161 591__ $$aTELECOMMUNICATIONS$$b28 / 120 = 0.233$$c2024$$dQ1$$eT1
000136161 593__ $$aComputer Science (miscellaneous)$$c2024$$dQ1
000136161 591__ $$aCOMPUTER SCIENCE, INFORMATION SYSTEMS$$b42 / 258 = 0.163$$c2024$$dQ1$$eT1
000136161 593__ $$aInformation Systems$$c2024$$dQ1
000136161 593__ $$aHuman-Computer Interaction$$c2024$$dQ1
000136161 593__ $$aComputer Science Applications$$c2024$$dQ1
000136161 655_4 $$ainfo:eu-repo/semantics/article$$vinfo:eu-repo/semantics/acceptedVersion
000136161 700__ $$aShahroodi, Taha
000136161 700__ $$aEscuin, Carlos
000136161 700__ $$aGaydadjiev, Georgi
000136161 700__ $$aWong, Stephan
000136161 700__ $$aHamdioui, Said
000136161 773__ $$g13, 2 (2024), 395 - 408$$pIEEE trans. emerg. top. comput.$$tIEEE Transactions on Emerging Topics in Computing$$x2168-6750
000136161 8564_ $$s14428042$$uhttps://zaguan.unizar.es/record/136161/files/texto_completo.pdf$$yPostprint
000136161 8564_ $$s3472675$$uhttps://zaguan.unizar.es/record/136161/files/texto_completo.jpg?subformat=icon$$xicon$$yPostprint
000136161 909CO $$ooai:zaguan.unizar.es:136161$$particulos$$pdriver
000136161 951__ $$a2025-11-13-15:00:46
000136161 980__ $$aARTICLE