Name
Title | ||
|---|---|---|
An Energy-Efficient Reconfigurable Processor for Binary-and Ternary-Weight Neural Networks With Flexible Data Bit Width |
Abstract | ||
|---|---|---|
Due to less memory requirement, low computation overhead and negligible accuracy degradation, deep neural networks with binary/ternary weights (BTNNs) have been widely employed on low-power mobile and Internet of Things (IoT) devices with limited storage capacity. Some hardware implementations have been proposed to accelerate the inference of BTNNs by utilizing the multiplication-free feature. However, some implicit characteristics in BTNN convolution, such as high arithmetic complexity and numerous redundant operations, are never considered. In this paper, we propose four optimization techniques to fully exploit these features. First, a feature-integral-based convolution (FIBC) method is proposed to reduce the arithmetic complexity of convolutional layers. Second, a kernel-transformation-feature-reconstruction (KTFR) convolution method is presented to remove redundant operations in BTNN convolution. Third, a hierarchical load-balancing mechanism (HLBM) is designed to eliminate zero value computation and improve resource utilization. Finally, a joint optimization approach for convolutional layers is proposed to search optimal calculation pattern for each layer. Based on the proposed four techniques, we design a reconfigurable processor in a 28-nm CMOS technology to accelerate the inferences of BTNNs. The four proposed techniques improve energy efficiency by 2.07
<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\times $ </tex-math></inline-formula>
, 1.65
<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\times $ </tex-math></inline-formula>
, 1.25
<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\times $ </tex-math></inline-formula>
, and 2.24
<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\times $ </tex-math></inline-formula>
for BTNNs respectively, compared with the baseline implementation which disables the proposed techniques. Benchmarked with binary-weight AlexNet, the processor achieves an energy efficiency of 19.9 TOPS/W at 200 MHz and 0.9 V. |
Year | DOI | Venue |
|---|---|---|
2019 | 10.1109/JSSC.2018.2881913 | IEEE Journal of Solid-State Circuits |
Keywords | Field | DocType |
Convolution,Kernel,Optimization,Neural networks,Hardware,Complexity theory,Standards | Kernel (linear algebra),Convolution,Computer science,Efficient energy use,Electronic engineering,CMOS,Ternary operation,Computational science,Artificial neural network,Binary number,Computation | Journal |
Volume | Issue | ISSN |
54 | 4 | 0018-9200 |
Citations | PageRank | References |
7 | 0.55 | 0 |
Authors | ||
7 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| shouyi yin | 1 | 579 | 99.95 |
| Peng Ouyang | 2 | 129 | 19.36 |
| Jianxun Yang | 3 | 12 | 2.80 |
| Tianyi Lu | 4 | 31 | 4.82 |
| Xiudong Li | 5 | 43 | 3.43 |
| leibo liu | 6 | 816 | 116.95 |
| Shaojun Wei | 7 | 555 | 102.32 |