Wei Chen

Dr. Wei Chen


Research Focus & Metrics

Academic Career Length

12Years (2015-Present)

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

Representative Publications

View All Publications
22 October 2024 in National Science Review
Verified
Journal Article
Open access
PDF file
YGYanchen Gui
Yanchen Gui
KWKai Wang
Kai Wang
ZJZhe Jin
Zhe Jin
HWHeyuan Wang
Heyuan Wang
HDHanzhi Deng
Hanzhi Deng
SPShilong Piao
Shilong Piao
Atmospheric CO2 growth rate (CGR), reflecting the carbon balance between anthropogenic emissions and net uptake from land and ocean, largely determines the magnitude and speed of global warming. CGR at Mauna Loa Baseline Observatory reached record high in 2023. Here we quantified major components of global carbon balance for 2023, by developing a framework which integrated fossil fuel CO2 emissions data, an atmospheric inversion from Global ObservatioN-based system for monitoring Greenhouse GAses (GONGGA) with two artificial intelligence (AI) models derived from dynamic global vegetation models. We attributed the record high CGR increase in 2023 compared to 2022 primarily to the large decline in land carbon sink (1803 ± 197 TgC year−1), with minor contributions from a small reduction in ocean carbon sink (184 TgC year−1) and a slight increase in fossil fuel emissions (24 TgC year−1). At least 78% of the global decline in land carbon sink was contributed by the decline in tropical sink, with GONGGA inversion (1354 TgC year−1) and AI simulations (1578 ± 666 TgC year−1) showing similar declines in the tropics. We further linked this tropical decline to the detrimental impact of El Niño-induced anomalous warming and drying on vegetation productivity in water-limited Sahel and southern Africa. Our successful attribution of CGR increase within the framework combining atmospheric inversion with AI simulations enabled near-real-time tracking of the global carbon budget which had a one-year reporting lag.

11 January 2022 in Frontiers in Public Health
Verified
Journal Article
Open access
YGYan Guo
Yan Guo
TWTengjiao Wang
Tengjiao Wang
WCWei Chen
Wei Chen
TKTed J. Kaptchuk
Ted J. Kaptchuk
XLXilian Li
Xilian Li
ZXZiming Xu
Ziming Xu
In the past decades, numerous clinical researches have been conducted to illuminate the effects of traditional Chinese medicine for better inheritance and promotion of it, which are mostly clinical trials designed from the doctor's point of view. This large-scale data mining study was conducted from real-world point of view in up to 10 years' big data sets of Traditional Chinese Medicine (TCM) in China, including both medical visits to hospital and cyberspace and contemporaneous social survey data. Finally, some important and interesting findings appear: (1) More Criticisms vs. More Visits. The intensity of criticism increased by 2.33 times over the past 10 years, while the actual number of visits increased by 2.41 times. (2) The people of younger age, highly educated and from economically developed areas have become the primary population for utilizing TCM, which is contrary to common opinions on the characteristics of TCM users. The discovery of this phenomenon indicates that TCM deserves further study on how it treats illness and maintains health.

12 March 2024 in IEEE Transactions on Knowledge and Data Engineering
Verified
Journal Article
Closed access
HWHeyuan Wang
Heyuan Wang
TWTengjiao Wang
Tengjiao Wang
SLShun Li
Shun Li
JZJiayi Zheng
Jiayi Zheng
WCWeijun Chen
Weijun Chen
WCWei Chen
Wei Chen
Stock forecast is a crucial yet challenging task in modern quantitative trading. Given theoretical and investment merits, recently a variety of deep learning methods have been proposed for automatically simulating stock movements from historical time series. However, these methods typically follow the i.i.d. assumption that actually contradicts the complex trading environment. In reality, individual stocks often exhibit diverse volatility patterns, while macro market scenarios may also change over time, jointly resulting in distribution shifts and weak generalization. To combat these bottlenecks, in this paper we propose a new learning architecture called Personalized Temporal Embedding and Routing ( PTER ) to improve stock forecast by forming a relaxed weight-sharing paradigm. The key of PTER is introducing hypernetworks to guide tailoring target network parameters, such that stock time series are embedded adapting to multi-object multi-scenario data disparities. Specifically, in the encoding stage, PTER first captures hyper-knowledge characterizing the similarity and peculiarity of different stocks and market scenarios. The knowledge space is then projected onto the temporal parameter space, enabling the customization of protruded features from chaotic observation signals. In the inference stage, each sample is dispatched to orthogonal predictor heads to dynamically output expected returns based on market conditions. Through experiments on benchmark datasets spanning over five years on four of the world's largest exchange markets, we show that PTER improves the cumulative and risk-adjusted revenue performance by a significant margin.

31 August 2022 in IEEE/ACM Transactions on Audio, Speech, and Language Processing
Verified
Journal Article
Closed access
YXYutao Xie
Yutao Xie
QWQiyu Wu
Qiyu Wu
WCWei Chen
Wei Chen
TWTengjiao Wang
Tengjiao Wang
Learning semantic sentence embeddings is beneficial to a variety of natural language processing tasks. Recently, methods using the contrastive learning framework to fine-tune pre-trained language models have been proposed and have achieved significant performance on sentence embeddings. However, sentence embeddings are easy to “overfit” to the contrastive learning goal. With the training of contrastive learning, the gap between contrastive learning and test tasks leads to unstable even declining performance on test tasks. For this reason, existing methods rely on the labeled development set to frequently evaluate the performance on test tasks and get the best checkpoints. In such a way, models are limited when the labeled data is unavailable or extremely scarce. To address this problem, we propose P seudo- S iamese network M utual L earning (PSML) for self-supervised sentence embeddings to reduce the gap between contrastive learning and test tasks. Consisting of the main encoder and the auxiliary encoder, PSML utilizes mutual learning as the basic framework. Between the two encoders, two mutual learning losses are constructed to share learning signals. The proposed model framework and losses of PSML help the model be optimized more stably and generalize better to test tasks, such as semantic textual similarity. Extensive experiments on seven public semantic textual similarity datasets show that PSML performs better than previous unsupervised contrastive methods for sentence embeddings. Besides, PSML also gives a stable performance curve on test tasks with training and is able to get the comparative performance without frequent evaluation on the labeled development set.

6 June 2025 in EBioMedicine
Verified
Journal Article
Open access
JCJuan Chen
Juan Chen
YJYao Jiahui
Yao Jiahui
WCWei Chen
Wei Chen
FZFeng Zhang
Feng Zhang
HWHeyuan Wang
Heyuan Wang
LZLan Zhu
Lan Zhu
Pelvic floor surface electromyography (sEMG) is widely used to evaluate and treat pelvic floor dysfunctions (PFDs). Based on sEMG, the Glazer protocol was developed over 20 years ago with a limited sample size, making it challenging to accurately diagnose PFDs across diverse populations and conditions. This study aims to establish a multidimensional database for monitoring pelvic floor sEMG, derive more reasonable reference ranges for sEMG parameters, and achieve accurate diagnosis of PFDs through artificial intelligence (AI). In this population-based, multicenter, cross-sectional study, we recruited 1605 participants from 21 centres across China, collected pelvic floor sEMG data, and established a multidimensional sEMG database. Based on the database, we developed an AI-Diagnostician-PFD diagnostic model, which leverages AI to derive AI-Reference ranges for sEMG parameters and diagnose PFDs. Data from 15 centres were divided into a training dataset (60%) and a test dataset (40%), while data from 6 additional centres were used to form an independent validation dataset. The proportions of normal and abnormal samples were consistent across the 15 and 6 centres, ensuring balanced representation. Additionally, both datasets encompassed diverse geographical regions, enhancing the model's generalizability. The diagnostic performance of the AI-Diagnostician-PFD model was evaluated on both the internal test dataset and the external validation dataset. In the external validation dataset, the area under the receiver operating characteristic curve (AUC) of the AI-Reference ranges was 11% higher than that of the Glazer standard. Specifically, the AUC values for the AI-Reference ranges on the internal and external validation datasets were 0.81 (95% CI: 8.13 × 10-1, 8.16 × 10-1) and 0.79 (95% CI: 7.90 × 10-1, 7.94 × 10-1), respectively, surpassing the AUC values of the Glazer standard of 0.76 (95% CI: 7.56 × 10-1, 7.59 × 10-1) and 0.68 (95% CI: 6.74 × 10-1, 6.78 × 10-1). Furthermore, the AI-Diagnostician-PFD model demonstrated superior diagnostic performance for PFDs, achieving an AUC 1% higher than other classical machine learning and deep learning models. The performance of the reference interval derived by AI surpassed that of the Glazer standard. Upon publication of this study, the AI-Diagnostician-PFD model for PFD prediction will be provided free via software on machines. Implementing this algorithm in clinical practice can enhance individual PFD diagnosis and improve population-level health outcomes. This study was supported by grants from the National Key R&D Program of China: The Establishment of a Comprehensive Network for PFD Prevention, Rehabilitation, Pelvic Floor Surgery and Related Complications (2021YFC2701300), Perception and Analysis of the Situation of Major Infectious Disease Outbreaks Based on Internet Big Data (2021ZD0111202), Research on New Models for Forecasting Major Infectious Diseases and Policy Evaluation (2021ZD0111205); Beijing Natural Science Foundation (7212073), National High-Level Hospital Clinical Research Funding (2022-PUMCH-B-087) and the Chinese Academy of Medical Sciences (CAMS) Innovation Fund for Medical Sciences (2021-I2M- C&T-B- 021).

12 April 2025 in Neural Networks
Verified
Journal Article
Closed access
MGMeng Gao
Meng Gao
YXYutao Xie
Yutao Xie
WCWei Chen
Wei Chen
FZFeng Zhang
Feng Zhang
FDFei Ding
Fei Ding
KWKam-Fai Wong
Kam-Fai Wong