Author name cluster

Genlang Chen

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

3 papers

2 author rows

EAAI Journal 2026 Journal Article

OPHEP-Miner: One-phase high-efficiency pattern mining utilizing tree structures

Xuan Liu
Genlang Chen
Fangyu Wu
Wanli Zuo
Youxi Wu

High-efficiency pattern mining aims to identify patterns generating substantial utility while minimizing costs. The existing mining methods mainly employ the pattern combination strategy or projection technology to generate high-order patterns, thereby resulting in a large number of candidate patterns. Although the generation of invalid patterns can be avoided by storing the transaction dataset in a tree structure for mining, this approach is considered time-consuming. To address this issue, we employed the prefix tree and utility vector structure to discover high-efficiency patterns within one phase. To further enhance mining efficiency, a bottom-up single-path pattern enumeration method was designed, which eliminates the need for complex tree structure constructions. Additionally, three pruning strategies were devised to limit the pattern search space, including a novel pruning technique specifically designed for single-path scenarios. The effectiveness of the proposed algorithm has been validated through comparative analyses conducted on eight diverse datasets using five algorithms. The results show that the proposed algorithm outperforms other approaches, particularly due to its effective pruning techniques, which enable it to run up to 64 times faster than the best competing algorithm in some cases. Furthermore, we used a movie dataset as a case study to illustrate the applicability of our proposed approach.

Details DOI

ICLR Conference 2025 Conference Paper

Dataset Ownership Verification in Contrastive Pre-trained Models

Yuechen Xie
Jie Song
Mengqi Xue
Haofei Zhang
Xingen Wang
Bingde Hu
Genlang Chen
Mingli Song

High-quality open-source datasets, which necessitate substantial efforts for curation, has become the primary catalyst for the swift progress of deep learning. Concurrently, protecting these datasets is paramount for the well-being of the data owner. Dataset ownership verification emerges as a crucial method in this domain, but existing approaches are often limited to supervised models and cannot be directly extended to increasingly popular unsupervised pre-trained models. In this work, we propose the first dataset ownership verification method tailored specifically for self-supervised pre-trained models by contrastive learning. Its primary objective is to ascertain whether a suspicious black-box backbone has been pre-trained on a specific unlabeled dataset, aiding dataset owners in upholding their rights. The proposed approach is motivated by our empirical insights that when models are trained with the target dataset, the unary and binary instance relationships within the embedding space exhibit significant variations compared to models trained without the target dataset. We validate the efficacy of this approach across multiple contrastive pre-trained models including SimCLR, BYOL, SimSiam, MOCO v3, and DINO. The results demonstrate that our method rejects the null hypothesis with a $p$-value markedly below $0.05$, surpassing all previous methodologies. Our code is available at https://github.com/xieyc99/DOV4CL.

Details

EAAI Journal 2022 Journal Article

Effective algorithms to mine skyline frequent-utility itemsets

Xuan Liu
Genlang Chen
Wanli Zuo

Skyline frequent-utility itemset mining is used to discover itemsets that are non-dominated by considering both support and utility factors. It is an extension of high-utility itemset mining. Most existing algorithms are based on the utility-list structure to mine skyline frequent-utility itemsets. A major limitation of utility-list based algorithms is that numerous join operations consume a huge amount of time and memory. To address this issue, two algorithms named EMSFUI-D and EMSFUI-B are proposed to mine skyline frequent-utility itemsets. EMSFUI-D performs the depth-first search to explore the search space of all itemsets. EMSFUI-B discovers itemsets based on the breadth-first search. Both algorithms utilize two pruning strategies to limit the search space. Moreover, in order to further facilitate the mining performance, the ISU-1 and ISU-2 structures are presented in EMSFUI-D to provide tighter utility upper bounds. These structures maintain the support and utility information of all 1-itemsets and 2-itemsets, respectively. Thus, there is no need to use these structures to prune search space in the breadth-first search algorithm. An extensive experimental study on real and synthetic datasets shows that our proposed algorithms outperform the state-of-the-art SKYFUP-D and SKYFUP-B algorithms in terms of execution time, memory consumption and pruning performance. Moreover, our designed algorithms are scalable for handling a large number of distinct items and transactions.

Details DOI