Shihui Yin

Shihui Yin

Auto-generated profile, based on publicly available information

Shihui Yin

School of Electrical, Computer and Energy Engineering, Arizona State University, Tempe, AZ, USA

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

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

Representative Publications

View All Publications
14 January 2020 in IEEE Journal of Solid-State Circuits
Verified
Journal Article
Closed access
SYShihui Yin
Shihui Yin
ZJZhewei Jiang
Zhewei Jiang
JSJae-Sun Seo
Jae-Sun Seo
MSMingoo Seok
Mingoo Seok

18 May 2020 in IEEE Journal of Solid-State Circuits
Verified
Journal Article
Closed access
ZJZhewei Jiang
Zhewei Jiang
SYShihui Yin
Shihui Yin
JSJae-Sun Seo
Jae-Sun Seo
MSMingoo Seok
Mingoo Seok
This article presents C3SRAM, an in-memory-computing SRAM macro. The macro is an SRAM module with the circuits embedded in bitcells and peripherals to perform hardware acceleration for neural networks with binarized weights and activations. The macro utilizes analog-mixed-signal (AMS) capacitive-coupling computing to evaluate the main computations of binary neural networks, binary-multiply-and-accumulate operations. Without the need to access the stored weights by individual row, the macro asserts all its rows simultaneously and forms an analog voltage at the read bitline node through capacitive voltage division. With one analog-to-digital converter (ADC) per column, the macro realizes fully parallel vector-matrix multiplication in a single cycle. The network type that the macro supports and the computing mechanism it utilizes are determined by the robustness and error tolerance necessary in AMS computing. The C3SRAM macro is prototyped in a 65-nm CMOS. It demonstrates an energy efficiency of 672 TOPS/W and a speed of 1638 GOPS (20.2 TOPS/mm²), achieving 3975x better energy-delay product than the conventional digital baseline performing the same operation. The macro achieves 98.3% accuracy for MNIST and 85.5% for CIFAR-10, which is among the best in-memory computing works in terms of energy efficiency and inference accuracy tradeoff.

19 August 2020 in IEEE Transactions on Electron Devices
Verified
Journal Article
Closed access
SYShihui Yin
Shihui Yin
XSXiaoyu Sun
Xiaoyu Sun
SYShimeng Yu
Shimeng Yu
JSJae-Sun Seo
Jae-Sun Seo
Deep neural network (DNN) hardware designs have been bottlenecked by conventional memories, such as SRAM due to density, leakage, and parallel computing challenges. Resistive devices can address the density and volatility issues but have been limited by peripheral circuit integration. In this work, we present a resistive RAM (RRAM)-based in-memory computing (IMC) design, which is fabricated in 90-nm CMOS with monolithic integration of RRAM devices. We integrated a 128x 64 RRAM array with CMOS peripheral circuits, including row/column decoders and flash analog-to-digital converters (ADCs), which collectively become a core component for scalable RRAM-based IMC for large DNNs. To maximize IMC parallelism, we assert all 128 wordlines of the RRAM array simultaneously, perform analog computing along the bitlines, and digitize the bitline voltages using ADCs. The resistance distribution of low-resistance states is tightened by an iterative write-verify scheme. Prototype chip measurements demonstrate high binary DNN accuracy of 98.5% for MNIST and 83.5% for CIFAR-10 data sets, with 24 TOPS/W and 158 GOPS. This represents 22.3x and 10.1x improvements in throughput and energy-delay product (EDP), respectively, compared with the state-of-the-art literature, which can enable intelligent functionalities for area-/energy-constrained edge computing devices.

19 October 2022 in IEEE Journal of Solid-State Circuits
Verified
Journal Article
Closed access
BZBo Zhang
Bo Zhang
SYShihui Yin
Shihui Yin
MKMinkyu Kim
Minkyu Kim
JSJyotishman Saikia
Jyotishman Saikia
SKSoonwan Kwon
Soonwan Kwon
MSMingoo Seok
Mingoo Seok
This article presents a programmable in-memory computing accelerator (PIMCA) for low-precision (1–2 b) deep neural network (DNN) inference. The custom 10T1C bitcell in the in-memory computing (IMC) macro has four additional transistors and one capacitor to perform capacitive-coupling-based multiply and accumulation (MAC) in analog-mixed-signal (AMS) domain. A macro containing $256\ttimes 128$ bitcells can simultaneously activate all the rows, and as a result, it can perform a matrix-vector multiplication (VMM) in one cycle. PIMCA integrates 108 of such IMC static random-access memory (SRAM) macros with the custom six-stage pipeline and the custom instruction set architecture (ISA) for instruction-level programmability. The results of IMC macros are fed to a single-instruction-multiple-data (SIMD) processor for other computations such as partial sum accumulation, max-pooling, activation functions, etc. To effectively use the IMC and SIMD datapath, we customize the ISA especially by adding hardware loop support, which reduces the program size by up to 73%. The accelerator is prototyped in a 28-nm technology, and integrates a total of 3.4-Mb IMC SRAM and 1.5-Mb off-the-shelf activation SRAM, demonstrating one of the largest IMC accelerators to date. It achieves the system-level energy efficiency of 437 TOPS/W and the peak throughput of 49 TOPS at the 42-MHz clock frequency and 1-V supply for the VGG9 and the ResNet-18 on the CIFAR-10 dataset.

13 May 2019 in IEEE Journal of Solid-State Circuits
Verified
Journal Article
Closed access
SYShihui Yin
Shihui Yin
MKMinkyu Kim
Minkyu Kim
DKDeepak Kadetotad
Deepak Kadetotad
YLYang Liu
Yang Liu
CBChisung Bae
Chisung Bae
JSJae-Sun Seo
Jae-Sun Seo
Many wearable devices employ the sensors for physiological signals (e.g., electrocardiogram or ECG) to continuously monitor personal health (e.g., cardiac monitoring). Considering private medical data storage, secure access to such wearable devices becomes a crucial necessity. Exploiting the ECG sensors present on wearable devices, we investigate the possibility of using ECG as the individually unique source for device authentication. In particular, we propose to use ECG features toward both cardiac monitoring and neural-network-based biometric authentication. For such complex functionalities to be seamlessly integrated in wearable devices, an accurate algorithm must be implemented with ultralow power and a small form factor. In this paper, a smart ECG processor is presented for ECG-based authentication as well as cardiac monitoring. Data-driven Lasso regression and low-precision techniques are developed to compress neural networks for feature extraction by 24.4x. The 65-nm testchip consumes 1.06 μW at 0.55 V for real-time ECG authentication. For authentication, equal error rates of 1.70%/2.18%/2.48% (best/average/worst) are achieved on the in-house 645-subject database. For cardiac monitoring, 93.13% arrhythmia detection sensitivity and 89.78% specificity are achieved for 42 subjects in the MIT-BIH arrhythmia database.

18 May 2020 in IEEE Journal of Solid-State Circuits
Verified
Journal Article
Closed access
DKDeepak Kadetotad
Deepak Kadetotad
SYShihui Yin
Shihui Yin
VBVisar Berisha
Visar Berisha
CCChaitali Chakrabarti
Chaitali Chakrabarti
JSJae-Sun Seo
Jae-Sun Seo
Long short-term memory (LSTM) is a type of recurrent neural networks (RNNs), which is widely used for time-series data and speech applications, due to its high accuracy on such tasks. However, LSTMs pose difficulties for efficient hardware implementation because they require a large amount of weight storage and exhibit computation complexity. Prior works have proposed compression techniques to alleviate the storage/computation requirements of LSTMs but elementwise sparsity schemes incur sizable index memory overhead and structured compression techniques report limited compression ratios. In this article, we present an energy-efficient LSTM RNN accelerator, featuring an algorithm-hardware co-optimized memory compression technique called hierarchical coarse-grain sparsity (HCGS). Aided by the HCGS-based blockwise recursive weight compression, we demonstrate LSTM networks with up to 16x fewer weights while achieving minimal error rate degradation. The prototype chip fabricated in 65-nm LP CMOS achieves up to 8.93 TOPS/W for real-time speech recognition using compressed LSTMs based on HCGS. HCGS-based LSTMs have demonstrated energy-efficient speech recognition with low error rates for TIMIT, TED-LIUM, and LibriSpeech data sets.