Accelerating artificial neural network computations by skipping input values
作者： Delerse Sebastien , Chouta Taoufik , Larzul Ludovic , Vangel Benoit Chappet De
DOI:
关键词:
摘要: Systems and methods for accelerating artificial neural network computation are disclosed. An example may comprise selecting, by a controller communicatively coupled to selector an arithmetic unit based on criterion, input value from the stream of values neuron, configuring, controller, provide, dynamically, selected unit, providing, information value, acquiring, information, weight set weights, performing, mathematical operation obtain result, wherein result is be used compute output neuron. The criterion include comparison between reference value. zero.
lens.org 参考文章(10)
Shay-Ping T. Wang, Artificial neuron using adder circuit and method of using same ,(1994)
Ravi Narayanaswami, Dong Hyuk Woo, Exploiting input data sparsity in neural network compute units ,(2016)
Mark Rollins, Arthur T. G. Fuller, Least mean squares adaptation of a concurrent multi-band pre-distorter using overlapping spines ,(2015)
Sanroku Tsukamoto, David Thach, Hirotaka Tamura, INFORMATION PROCESSING APPARATUS, ISING DEVICE, AND INFORMATION PROCESSING APPARATUS CONTROL METHOD ,(2017)
Ould-Ahmed-Vall Elmoustapha, Brown William M, Systems, apparatuses, and methods for cumulative product ,(2018)
Haiut Moshe, NEURAL CELL AND A NEURAL NETWORK ,(2018)
William J. Dally, Franciscus Wilhelmus Sijstermans, Jeffrey Michael Pool, Xiaojun Wang, Liang Chen, Zhou Yan, Yuanzhi Hua, Data compaction and memory bandwidth reduction for sparse neural networks ,(2018)
Henry G Glenn, Parks Terry, NEURAL NETWORK UNIT THAT INTERRUPTS PROCESSING CORE UPON CONDITION ,(2018)
Kalsi Gurpreet S, Mishra Amit, Pillai Kamlesh, DEEP NEURAL NETWORK ARCHITECTURE USING PIECEWISE LINEAR APPROXIMATION ,(2020)
Martin Christopher, Neural Network Architecture Using Convolution Engine Filter Weight Buffers ,(2019)