Yifei Shen

Yifei Shen

Auto-generated profile, based on publicly available information

Yifei Shen

Telecommunications Circuits Laboratory (TCL), cole Polytechnique Fdrale de Lausanne (EPFL), Lausanne, Switzerland

Research Focus & Metrics

Academic Career Length

11Years (2016-Present)

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

Representative Publications

View All Publications
8 March 2024 in IEEE Transactions on Circuits and Systems I: Regular Papers
Verified
Journal Article
Closed access
YRYuqing Ren
Yuqing Ren
HHHassan Harb
Hassan Harb
YSYifei Shen
Yifei Shen
ABAlexios Balatsoukas-Stimming
Alexios Balatsoukas-Stimming
ABAndreas Burg
Andreas Burg
5G New Radio (NR) has stringent demands on both performance and complexity for the design of low-density parity-check (LDPC) decoding algorithms and corresponding VLSI implementations. Furthermore, decoders must fully support the wide range of all 5G NR blocklengths and code rates, which is a significant challenge. In this paper, we present a high-performance and low-complexity LDPC decoder, tailor-made to fulfill the 5G requirements. First, to close the gap between belief propagation (BP) decoding and its approximations in hardware, we propose an extension of adjusted min-sum decoding, called generalized adjusted min-sum (GA-MS) decoding. This decoding algorithm flexibly truncates the incoming messages at the check node level and carefully approximates the non-linear functions of BP decoding to balance the error-rate and hardware complexity. Numerical results demonstrate that the proposed fixed-point GA-MS has only a minor gap of 0.1 dB compared to floating-point BP under various scenarios of 5G standard specifications. Secondly, we present a fully reconfigurable 5G NR LDPC decoder implementation based on GA-MS decoding. Given that memory occupies a substantial portion of the decoder area, we adopt multiple data compression and approximation techniques to reduce 42.2% of the memory overhead. The corresponding 28nm FD-SOI ASIC decoder has a core area of 1.823 mm $^{2}$ and operates at 895 MHz. It is compatible with all 5G NR LDPC codes and achieves a peak throughput of 24.42 Gbps and a maximum area efficiency of 13.40 Gbps/mm $^{2}$ at 4 decoding iterations.

14 February 2022 in IEEE Transactions on Signal Processing
Verified
Journal Article
Closed access
YSYifei Shen
Yifei Shen
WZWenyue Zhou
Wenyue Zhou
YHYongming Huang
Yongming Huang
ZZZaichen Zhang
Zaichen Zhang
XYXiaohu You
Xiaohu You
CZChuan Zhang
Chuan Zhang
Iterative detection and decoding (IDD) enjoys a higher capacity than separate detection and decoding (SDD). However, IDD requires both the detector and the decoder to support soft input and soft output. Considering the standard channel codes in 5G communication systems, the preferred decoder for polar codes, the successive cancellation list (SCL) decoder, outputs only hard decisions on the side of codeword. Yet, current soft-output polar decoders, such as belief propagation (BP) and soft cancellation (SCAN), are much inferior to the SCL decoders in terms of error correction performance. To this end, we propose a soft-output list (SOL) decoder in this paper, which considers both hypotheses of 0 and 1 for unreliable bits and keeps the a-priori likelihoods for reliable bits, so that it can provide soft output and satisfactory error-rate performance at the same time. Exploiting the property of special nodes, the resulting FastSOL decoder can directly return soft messages from these nodes, thus speeding up the decoding. The error-rate performance of the FastSOL decoder enhances as the number of iterations increases in the non-IDD setup. The proposed FastSOL decoder can also be applied in the IDD system. Concatenated with a maximum a-posteriori detector in ten IDD loops, our decoder with a list size of eight exhibits over 1 dB gain compared to using the SCAN decoder for polar codes with length 256 and half rate, which also outperforms the state-of-the-art soft list decoder by 0.12 dB in the same settings with a higher computational complexity, but comparable decoding latency.

1 February 2024 in IEEE Transactions on Signal Processing
Verified
Journal Article
Closed access
YRYuqing Ren
Yuqing Ren
YSYifei Shen
Yifei Shen
LZLeyu Zhang
Leyu Zhang
AKAndreas Toftegaard Kristensen
Andreas Toftegaard Kristensen
ABAlexios Balatsoukas-Stimming
Alexios Balatsoukas-Stimming
CZChuan Zhang
Chuan Zhang
Due to its high parallelism, belief propagation (BP) decoding is amenable to high-throughput applications and thus represents a promising solution for the ultra-high peak data rate required by future communication systems. To bridge the performance gap compared to the widely used successive cancellation list (SCL) decoding algorithm, BP list (BPL) decoding for polar codes extends candidate codeword exploration via multiple permuted factor graphs (PFGs) to improve the error-correcting performance of BP decoding. However, it is a significant challenge to design a unified and flexible BPL hardware architecture that supports various PFGs and code configurations. In this paper, we present the first VLSI implementation of a BPL decoder for polar codes that overcomes this implementation challenge with a hardware-friendly algorithm for on-the-fly flexible permutations. First, we introduce a sequential generation (SG) algorithm to obtain a near-optimal PFG set. Additionally, we demonstrate that any permutation can be decomposed into a combination of multiple fixed routings, and design a low-complexity permutation network to generate graphs in an on-the-fly fashion. Our BPL decoder has a low decoding latency by executing decoding and permutation generation in parallel and supports arbitrary list sizes without area overhead. Experimental results based on 28nm FD-SOI technology show that for length-1024 polar codes with a code rate of one-half, our BPL decoder with 32 PFGs exhibits similar error-correcting performance to SCL with a list size of 4 and achieves an average throughput of 25.63 Gbps and an area efficiency of 29.46 Gbps/mm 2 , which is 1.82× and 4.33× faster than the state-of-the-art BP flip and SCL decoders, respectively

22 November 2021 in IEEE Wireless Communications Letters
Verified
Journal Article
Closed access
YSYifei Shen
Yifei Shen
ABAlexios Balatsoukas-Stimming
Alexios Balatsoukas-Stimming
XYXiaohu You
Xiaohu You
CZChuan Zhang
Chuan Zhang
ABAndreas Peter Burg
Andreas Peter Burg
Since polar were ratified as part of the 5G standard, low-complexity polar decoders with close-to-optimum error-rate performance have received significant attention. Compared to successive cancellation (SC) decoding, both SC list and SC flip decoding can improve the error rate performance by increasing the number of considered candidate solutions. The combination of both strategies leads to SC list flip (SCLF) decoding, which can provide a tradeoff between error rate performance and area as well as energy. In this letter, we derive a new flip metric for the SCLF decoding process, based on which we propose the dynamic SCLF (D-SCLF) decoding algorithm. Moreover, we exploit the distributed CRC defined in the 5G standard to further optimize the D-SCLF decoding. Numerical results show that for the downlink control channel, our D-SCLF decoder with a list size of only four and only three additional attempts can achieve the performance of a regular list decoder with a list size of eight, leading to an overall memory and average complexity (energy) reduction.

24 August 2022 in Scientific Reports
Verified
Journal Article
Open access
PDF file
BYBo Ye
Bo Ye
YSYifei Shen
Yifei Shen
HCHui Chen
Hui Chen
SLSha Lin
Sha Lin
WMweilin mao
weilin mao
XLXuefen Li
Xuefen Li
Hepatitis B virus (HBV) infection is still a major public health problem worldwide. We aimed to identify new, non-invasive biomarkers for the early diagnosis of chronic HBV-related diseases, reveal alterations in the progression of chronic hepatitis B (CHB), liver cirrhosis (LC), and hepatocellular carcinoma (HCC). Here, exosomes were isolated and characterized through size exclusion chromatography and nanoparticle tracking analysis. Profiles of differentially expressed proteins (DEPs) were analyzed through liquid chromatography-tandem mass spectrometry (LC–MS/MS), Gene Ontology, and Kyoto Encyclopedia of Genes and Genomes analyses. Results showed that the DEPs, including CO9, LBP, SVEP1, and VWF levels in extracellular vesicles (EVs) were significantly higher in CHB than in healthy controls (HCs). VWF expression levels in EVs were significantly lower in CHB than in those with LC. KV311 expression levels in EVs were significantly higher, whereas LBP levels were significantly lower in patients with CHB than in those with HCC. All biomarkers seemed to exhibit a high diagnostic capacity for HBV-related liver disease. Patients with HBV-induced chronic liver disease exhibit characteristic protein profiles in their EVs. Thus, serum exosomes may be used as novel, liquid biopsy biomarkers to provide useful clinical information for the diagnosis of HBV-related liver diseases at different stages.

16 May 2022
Verified
Conference Paper
Open access
YSYifei Shen
Yifei Shen
JZJun Zhang
Jun Zhang
KLKhaled B. Letaief
Khaled B. Letaief
In recent years, there has been a surge in applying deep learning to various challenging design problems in communication networks. The early attempts adopt neural architectures inherited from applications such as computer vision, which suffer from poor generalization, scalability, and lack of interpretability. To tackle these issues, domain knowledge has been integrated into the neural architecture design, which achieves near-optimal performance in large-scale networks and generalizes well under different system settings. This paper endeavors to theoretically validate the importance and effects of neural architectures when applying deep learning to communication network design. We prove that by exploiting permutation invariance, a common property in communication networks, graph neural networks (GNNs) converge faster and generalize better than fully connected multi-layer perceptrons (MLPs), especially when the number of nodes (e.g., users, base stations, or antennas) is large. Specifically, we prove that under common assumptions, for a communication network with n nodes, GNNs converge O(n log n) times faster and their generalization error is O(n) times lower, compared with MLPs.