Peilin Zhou

Peilin Zhou

Auto-generated profile, based on publicly available information

Peilin Zhou

New York University, Abu Dhabi, Abu Dhabi, United Arab Emirates

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

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

Representative Publications

View All Publications
7 April 2025 in Nature Reviews Bioengineering
Verified
Journal Article
Closed access
PDF file
FLFenglin Liu
Fenglin Liu
HZHongjian Zhou
Hongjian Zhou
BGBoyang Gu
Boyang Gu
XZXinyu Zou
Xinyu Zou
JHJinfa Huang
Jinfa Huang
DCDavid Clifton
David Clifton

14 September 2023
Verified
Conference Paper
Open access
YXYueqi Xie
Yueqi Xie
JGJingqi Gao
Jingqi Gao
PZPeilin Zhou
Peilin Zhou
QYQichen Ye
Qichen Ye
YHYining Hua
Yining Hua
SKSung Hun Kim
Sung Hun Kim
Existing research efforts for multi-interest candidate matching in recommender systems mainly focus on improving model architecture or incorporating additional information, neglecting the importance of training schemes. This work revisits the training framework and uncovers two major problems hindering the expressiveness of learned multi-interest representations. First, the current training objective (i.e., uniformly sampled softmax) fails to effectively train discriminative representations in a multi-interest learning scenario due to the severe increase in easy negative samples. Second, a routing collapse problem is observed where each learned interest may collapse to express information only from a single item, resulting in information loss. To address these issues, we propose the REMI framework, consisting of an Interest-aware Hard Negative mining strategy (IHN) and a Routing Regularization (RR) method. IHN emphasizes interest-aware hard negatives by proposing an ideal sampling distribution and developing a Monte-Carlo strategy for efficient approximation. RR prevents routing collapse by introducing a novel regularization term on the item-to-interest routing matrices. These two components enhance the learned multi-interest representations from both the optimization objective and the composition information. REMI is a general framework that can be readily applied to various existing multi-interest candidate matching methods. Experiments on three real-world datasets show our method can significantly improve state-of-the-art methods with easy implementation and negligible computational overhead. The source code is available at https://github.com/Tokkiu/REMI.

21 October 2023
Verified
Conference Paper
Open access
PZPeilin Zhou
Peilin Zhou
QYQichen Ye
Qichen Ye
YXYueqi Xie
Yueqi Xie
JGJingqi Gao
Jingqi Gao
Shoujin WangShoujin Wang
Shoujin Wang
SKSung Hun Kim
Sung Hun Kim
Transformer-based sequential recommendation (SR) has been booming in recent years, with the self-attention mechanism as its key component. Self-attention has been widely believed to be able to effectively select those informative and relevant items from a sequence of interacted items for next-item prediction via learning larger attention weights for these items. However, this may not always be true in reality. Our empirical analysis of some representative Transformer-based SR models reveals that it is not uncommon for large attention weights to be assigned to less relevant items, which can result in inaccurate recommendations. Through further in-depth analysis, we find two factors that may contribute to such inaccurate assignment of attention weights:sub-optimal position encoding andnoisy input. To this end, in this paper, we aim to address this significant yet challenging gap in existing works. To be specific, we propose a simple yet effective framework called Attention Calibration for Transformer-based Sequential Recommendation (AC-TSR). In AC-TSR, a novel spatial calibrator and adversarial calibrator are designed respectively to directly calibrates those incorrectly assigned attention weights. The former is devised to explicitly capture the spatial relationships (i.e., order and distance) among items for more precise calculation of attention weights. The latter aims to redistribute the attention weights based on each item's contribution to the next-item prediction. AC-TSR is readily adaptable and can be seamlessly integrated into various existing transformer-based SR models. Extensive experimental results on four benchmark real-world datasets demonstrate the superiority of our proposed AC-TSR via significant recommendation performance enhancements. The source code is available at https://github.com/AIM-SE/AC-TSR.

14 September 2023
Verified
Conference Paper
Open access
PZPeilin Zhou
Peilin Zhou
JGJingqi Gao
Jingqi Gao
YXYueqi Xie
Yueqi Xie
QYQichen Ye
Qichen Ye
YHYining Hua
Yining Hua
SKSung Hun Kim
Sung Hun Kim

13 May 2024
Verified
Conference Paper
Open access
PZPeilin Zhou
Peilin Zhou
YHYou-Liang Huang
You-Liang Huang
YXYueqi Xie
Yueqi Xie
JGJingqi Gao
Jingqi Gao
Shoujin WangShoujin Wang
Shoujin Wang
SKSung Hun Kim
Sung Hun Kim
Sequential recommender systems (SRS) are designed to predict users' future behaviors based on their historical interaction data. Recent research has increasingly utilized contrastive learning (CL) to leverage unsupervised signals to alleviate the data sparsity issue in SRS. In general, CL-based SRS first augments the raw sequential interaction data by using data augmentation strategies and employs a contrastive training scheme to enforce the representations of those sequences from the same raw interaction data to be similar. Despite the growing popularity of CL, data augmentation, as a basic component of CL, has not received sufficient attention. This raises the question: Is it possible to achieve superior recommendation results solely through data augmentation? To answer this question, we benchmark eight widely used data augmentation strategies, as well as state-of-the-art CL-based SRS methods, on four real-world datasets under both warm- and cold-start settings. Intriguingly, the conclusion drawn from our study is that, certain data augmentation strategies can achieve similar or even superior performance compared with some CL-based methods, demonstrating the potential to significantly alleviate the data sparsity issue with fewer computational overhead. We hope that our study can further inspire more fundamental studies on the key functional components of complex CL techniques. Our processed datasets and codes are available at https://github.com/AIM-SE/DA4Rec.

5 March 2025 in Journal of Medical Internet Research
Verified
Journal Article
Open access
WLWanxin Li
Wanxin Li
YHYining Hua
Yining Hua
PZPeilin Zhou
Peilin Zhou
LZLi Zhou
Li Zhou
XXXin Xu
Xin Xu
JYJie Yang
Jie Yang
Background: While the COVID-19 pandemic has induced massive discussion of available medications on social media, traditional studies focused only on limited aspects, such as public opinions, and endured reporting biases, inefficiency, and long collection times. Objective: Harnessing drug-related data posted on social media in real-time can offer insights into how the pandemic impacts drug use and monitor misinformation. This study aimed to develop a natural language processing (NLP) pipeline tailored for the analysis of social media discourse on COVID-19–related drugs. Methods: This study constructed a full pipeline for COVID-19–related drug tweet analysis, using pretrained language model–based NLP techniques as the backbone. This pipeline is architecturally composed of 4 core modules: named entity recognition and normalization to identify medical entities from relevant tweets and standardize them to uniform medication names for time trend analysis, target sentiment analysis to reveal sentiment polarities associated with the entities, topic modeling to understand underlying themes discussed by the population, and drug network analysis to dig potential adverse drug reactions (ADR) and drug-drug interactions (DDI). The pipeline was deployed to analyze tweets related to the COVID-19 pandemic and drug therapies between February 1, 2020, and April 30, 2022. Results: From a dataset comprising 169,659,956 COVID-19–related tweets from 103,682,686 users, our named entity recognition model identified 2,124,757 relevant tweets sourced from 1,800,372 unique users, and the top 5 most-discussed drugs: ivermectin, hydroxychloroquine, remdesivir, zinc, and vitamin D. Time trend analysis revealed that the public focused mostly on repurposed drugs (ie, hydroxychloroquine and ivermectin), and least on remdesivir, the only officially approved drug among the 5. Sentiment analysis of the top 5 most-discussed drugs revealed that public perception was predominantly shaped by celebrity endorsements, media hot spots, and governmental directives rather than empirical evidence of drug efficacy. Topic analysis obtained 15 general topics of overall drug-related tweets, with “clinical treatment effects of drugs” and “physical symptoms” emerging as the most frequently discussed topics. Co-occurrence matrices and complex network analysis further identified emerging patterns of DDI and ADR that could be critical for public health surveillance like better safeguarding public safety in medicines use. Conclusions: This study shows that an NLP-based pipeline can be a robust tool for large-scale public health monitoring and can offer valuable supplementary data for traditional epidemiological studies concerning DDI and ADR. The framework presented here aspires to serve as a cornerstone for future social media–based public health analytics.