Yizhi Wang

Dr. Yizhi Wang

orcid

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 February 2018 in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
Verified
Journal Article
Closed access
YWYizhi Wang
Yizhi Wang
JLJun Lin
Jun Lin
ZWZhongfeng Wang
Zhongfeng Wang
Binary weight convolutional neural networks (BCNNs) can achieve near state-of-the-art classification accuracy and have far less computation complexity compared with traditional CNNs using high-precision weights. Due to their binary weights, BCNNs are well suited for vision-based Internet-of-Things systems being sensitive to power consumption. BCNNs make it possible to achieve very high throughput with moderate power dissipation. In this paper, an energy-efficient architecture for BCNNs is proposed. It fully exploits the binary weights and other hardware-friendly characteristics of BCNNs. A judicious processing schedule is proposed so that off-chip I/O access is minimized and activations are maximally reused. To significantly reduce the critical path delay, we introduce optimized compressor trees and approximate binary multipliers with two novel compensation schemes. The latter is able to save significant hardware resource, and almost no computation accuracy is compromised. Taking advantage of error resiliency of BCNNs, an innovative approximate adder is developed, which significantly reduces the silicon area and data path delay. Thorough error analysis and extensive experimental results on several data sets show that the approximate adders in the data path cause negligible accuracy loss. Moreover, algorithmic transformations for certain layers of BCNNs and a memory-efficient quantization scheme are incorporated to further reduce the energy cost and on-chip storage requirement. Finally, the proposed BCNN hardware architecture is implemented with the SMIC 130-nm technology. The postlayout results demonstrate that our design can achieve an energy efficiency over 2.0TOp/s/W when scaled to 65 nm, which is more than two times better than the prior art.

1 January 2022 in Journal of Intelligent Systems
Verified
Journal Article
Open access
PDF file
ZLZihao Li
Zihao Li
YWYizhi Wang
Yizhi Wang
ZYZhong Yang
Zhong Yang
XTXiaomin Tian
Xiaomin Tian
LZLixin Zhai
Lixin Zhai
YZYang Zhang
Yang Zhang
Fingerprint recognition is the most widely used identification method at present. However, it still falls short in terms of cross-platform and algorithmic complexity, which exerts a certain effect on the migration of fingerprint data and the development of the system. The conventional image recognition methods require offline standard databases constructed in advance for image access efficiency. The database can provide a pre-processed image via a specific method that probably is compatible merely with the specific recognition algorithm. Then, the specific recognition algorithm starts the process of retrieving these specific pre-proessing images for recognition and inevitably will be blocked from other datasets. The proposed method in this research designed an embedded image processing algorithm based on a Siamese neural network in the recognition method that allows the proposed method to recognize images from any source without constructing a database for image storage in advance. In this research, the proposed method was applied to fingerprint recognition and evaluation of the proposed method was evaluated. The results showed that the accuracy of the proposed algorithm was up to 92%, and itsF1 score was up to 0.87. Compared with the conventional fingerprint matching methods, its significant advantage in the FRR, FAR, and CR jointly indicated the remarkable correct recognition rate of the proposed method.

27 July 2018 in IEEE Transactions on Circuits and Systems I: Regular Papers
Verified
Journal Article
Closed access
YWYizhi Wang
Yizhi Wang
JLJun Lin
Jun Lin
ZWZhongfeng Wang
Zhongfeng Wang
Emerging convolutional neural networks (CNNs) tend to be designed with varied per-layer data widths and sparse representations. However, these two features, which bring many redundant computations, have not been exploited simultaneously in existing hardware architectures for CNNs. This paper proposes an energy-quality scalable architecture, namely folded precision-adjustable processor (FPAP), to eliminate all computational redundancies by using folding techniques. On one hand, FPAP decomposes the dominant multiply-accumulate (MAC) operations into multiple adds and folds them into single arithmetic unit. Only effective adds (or part of them) are then calculated serially. Thus, FPAP can adapt to different per-layer data widths and enable precision-adjustable approximate computing. Particularly, FPAP adaptively selects either activation or weight to be decomposed in every single MAC to minimize the total number of adds and clock cycles. On the other hand, a 1-D convolution is undertaken by a multi-tap transposed finite impulse response (FIR) filter, which is folded into one tap to skip MACs with zero weights or activations. Besides, a judicious delay element remapping scheme and a novel genetic algorithm-based kernel reallocation scheme, are developed to reduce the power consumption in a folded FIR filter and mitigate the load imbalance issue caused by irregular sparsity, respectively. With all these optimizations, FPAP is able to reach comparable or even faster processing speed over the corresponding unfolded design in sparse CNNs while consuming smaller area. Experimental results on real CNN models demonstrate that FPAP can scale its energy efficiency from 4.28 to 23.63 TOP/s/W, and area efficiency from 37.79 to 164.15GOP/s/mm², respectively, under the TSMC 28-nm HPC CMOS technology.

1 January 2021 in Complexity
Verified
Journal Article
Open access
PDF file
SGShanshan Gu
Shanshan Gu
GXGuangrong Xi
Guangrong Xi
LGLingyu Ge
Lingyu Ge
TYTingting Yin
Tingting Yin
YWYizhi Wang
Yizhi Wang
ZYZhenzhong Yu
Zhenzhong Yu
A terahertz (THz) frequency‐modulated continuous wave (FMCW) imaging radar system is developed for high‐resolution 3D imaging recently. Aiming at the problems of long data acquisition periods and large sample sizes for the developed imaging system, an algorithm based on compressed sensing is proposed for THz FMCW radar 3D imaging in this paper. Firstly, the FMCW radar signal model is built, and the conventional range migration algorithm is introduced for THz FMCW radar imaging. Then, compressed sensing is extended for THz FMCW radar 3D imaging, and the Newton smooth L0‐norm (NSL0) algorithm is presented for sparse measurement data reconstruction. Both simulation and measurement experiments demonstrate the feasibility of reconstructing THz images from measurements even at the sparsity rate of 20%.

3 August 2022 in Journal of Sensors
Verified
Journal Article
Open access
PDF file
XHXingliu Hu
Xingliu Hu
HSHaifei Si
Haifei Si
JMJunhui Mao
Junhui Mao
YWYizhi Wang
Yizhi Wang
In this study, a new square-based fiber Bragg grating (FBG) sensor network model is proposed to address possible link failures in FBG sensor networks and improve their reliability. Graph theory and optical switching are simultaneously applied to these sensor networks to improve their self-healing ability; the FBG sensor network is regarded as a directed graph. Three commonly used self-short-circuit algorithms are compared in terms of the self-healing capabilities that they provide to the optical fiber sensor network. Among these, the shortest-path faster algorithm achieved a high, nearly 90% repair accuracy and had an average repair time of 0.103 s, the shortest in this study. The newly designed FBG self-healing network can be reorganized and repaired when local damage occurs, thereby improving its reliability.

13 December 2021 in Molecules
Verified
Journal Article
Open access
PDF file
LHLingyi Huang
Lingyi Huang
YZYu Zhou
Yu Zhou
YWYizhi Wang
Yizhi Wang
MLMin Lin
Min Lin
From unstable crystals to relatively stable monohydrate crystals, many researchers have been working on S-nitrosocaptopril for more than two decades. S-nitrosocaptopril monohydrate (Cap-NO·H2O) is a novel crystal form of S-nitrosocaptopril (Cap-NO), and is not only a nitric oxide (NO) donor, but also an angiotensin-converting enzyme inhibitor (ACEI). Yet, a method for long-term storage has never been reported. In order to determine the optimal storage conditions, Plackett–Burman (PB) design was performed to confirm the critical factors. Response surface methodology (RSM) was employed to determine the optimal Cap-NO·H2O storage condition, based on the rough interval determined by the path of steepest ascent experiment. The optimized storage condition was denoted as nitrogen purity of 97%, temperature of −10 °C and 1.20 g deoxidizer. In this case, a final preservation rate of 97.91 ± 0.59% could be obtained. In specific storage conditions, Cap-NO·H2O was found to be stable for at least 6 months in individual PE package, procreating a potentially applicable avenue.