A Sliding‐Kernel Computation‐In‐Memory Architecture for Convolutional Neural Network
Abstract Presently described is a sliding‐kernel computation‐in‐memory (SKCIM) architecture conceptually involving two overlapping layers of functional arrays, one containing memory elements and artificial synapses for neuromorphic computation, the other is used for storing and sliding convolutional...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2024-12-01
|
| Series: | Advanced Science |
| Subjects: | |
| Online Access: | https://doi.org/10.1002/advs.202407440 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850247177761193984 |
|---|---|
| author | Yushen Hu Xinying Xie Tengteng Lei Runxiao Shi Man Wong |
| author_facet | Yushen Hu Xinying Xie Tengteng Lei Runxiao Shi Man Wong |
| author_sort | Yushen Hu |
| collection | DOAJ |
| description | Abstract Presently described is a sliding‐kernel computation‐in‐memory (SKCIM) architecture conceptually involving two overlapping layers of functional arrays, one containing memory elements and artificial synapses for neuromorphic computation, the other is used for storing and sliding convolutional kernel matrices. A low‐temperature metal‐oxide thin‐film transistor (TFT) technology capable of monolithically integrating single‐gate TFTs, dual‐gate TFTs, and memory capacitors is deployed for the construction of a physical SKCIM system. Exhibiting an 88% reduction in memory access operations compared to state‐of‐the‐art systems, a 32 × 32 SKCIM system is applied to execute common convolution tasks. A more involved demonstration is the application of a 5‐layer, SKCIM‐based convolutional neural network to the classification of the modified national institute of standards and technology (MNIST) dataset of handwritten numerals, achieving an accuracy rate of over 95%. |
| format | Article |
| id | doaj-art-c4e1c44d2e694cb39749d844fc7832e3 |
| institution | OA Journals |
| issn | 2198-3844 |
| language | English |
| publishDate | 2024-12-01 |
| publisher | Wiley |
| record_format | Article |
| series | Advanced Science |
| spelling | doaj-art-c4e1c44d2e694cb39749d844fc7832e32025-08-20T01:59:00ZengWileyAdvanced Science2198-38442024-12-011146n/an/a10.1002/advs.202407440A Sliding‐Kernel Computation‐In‐Memory Architecture for Convolutional Neural NetworkYushen Hu0Xinying Xie1Tengteng Lei2Runxiao Shi3Man Wong4State Key Laboratory of Advanced Displays and Optoelectronics Technologies Department of Electronic and Computer Engineering The Hong Kong University of Science and Technology (HKUST) Hong Kong ChinaState Key Laboratory of Advanced Displays and Optoelectronics Technologies Department of Electronic and Computer Engineering The Hong Kong University of Science and Technology (HKUST) Hong Kong ChinaState Key Laboratory of Advanced Displays and Optoelectronics Technologies Department of Electronic and Computer Engineering The Hong Kong University of Science and Technology (HKUST) Hong Kong ChinaState Key Laboratory of Advanced Displays and Optoelectronics Technologies Department of Electronic and Computer Engineering The Hong Kong University of Science and Technology (HKUST) Hong Kong ChinaState Key Laboratory of Advanced Displays and Optoelectronics Technologies Department of Electronic and Computer Engineering The Hong Kong University of Science and Technology (HKUST) Hong Kong ChinaAbstract Presently described is a sliding‐kernel computation‐in‐memory (SKCIM) architecture conceptually involving two overlapping layers of functional arrays, one containing memory elements and artificial synapses for neuromorphic computation, the other is used for storing and sliding convolutional kernel matrices. A low‐temperature metal‐oxide thin‐film transistor (TFT) technology capable of monolithically integrating single‐gate TFTs, dual‐gate TFTs, and memory capacitors is deployed for the construction of a physical SKCIM system. Exhibiting an 88% reduction in memory access operations compared to state‐of‐the‐art systems, a 32 × 32 SKCIM system is applied to execute common convolution tasks. A more involved demonstration is the application of a 5‐layer, SKCIM‐based convolutional neural network to the classification of the modified national institute of standards and technology (MNIST) dataset of handwritten numerals, achieving an accuracy rate of over 95%.https://doi.org/10.1002/advs.202407440convolutional computingconvolutional neural networkmetal‐oxideneuromorphic computingthin film transistor |
| spellingShingle | Yushen Hu Xinying Xie Tengteng Lei Runxiao Shi Man Wong A Sliding‐Kernel Computation‐In‐Memory Architecture for Convolutional Neural Network Advanced Science convolutional computing convolutional neural network metal‐oxide neuromorphic computing thin film transistor |
| title | A Sliding‐Kernel Computation‐In‐Memory Architecture for Convolutional Neural Network |
| title_full | A Sliding‐Kernel Computation‐In‐Memory Architecture for Convolutional Neural Network |
| title_fullStr | A Sliding‐Kernel Computation‐In‐Memory Architecture for Convolutional Neural Network |
| title_full_unstemmed | A Sliding‐Kernel Computation‐In‐Memory Architecture for Convolutional Neural Network |
| title_short | A Sliding‐Kernel Computation‐In‐Memory Architecture for Convolutional Neural Network |
| title_sort | sliding kernel computation in memory architecture for convolutional neural network |
| topic | convolutional computing convolutional neural network metal‐oxide neuromorphic computing thin film transistor |
| url | https://doi.org/10.1002/advs.202407440 |
| work_keys_str_mv | AT yushenhu aslidingkernelcomputationinmemoryarchitectureforconvolutionalneuralnetwork AT xinyingxie aslidingkernelcomputationinmemoryarchitectureforconvolutionalneuralnetwork AT tengtenglei aslidingkernelcomputationinmemoryarchitectureforconvolutionalneuralnetwork AT runxiaoshi aslidingkernelcomputationinmemoryarchitectureforconvolutionalneuralnetwork AT manwong aslidingkernelcomputationinmemoryarchitectureforconvolutionalneuralnetwork AT yushenhu slidingkernelcomputationinmemoryarchitectureforconvolutionalneuralnetwork AT xinyingxie slidingkernelcomputationinmemoryarchitectureforconvolutionalneuralnetwork AT tengtenglei slidingkernelcomputationinmemoryarchitectureforconvolutionalneuralnetwork AT runxiaoshi slidingkernelcomputationinmemoryarchitectureforconvolutionalneuralnetwork AT manwong slidingkernelcomputationinmemoryarchitectureforconvolutionalneuralnetwork |