Scholarlink

<p>To address the challenges of high user involvement and low time efficiency in handling large datasets, we propose an enhanced interactive clustering algorithm called LR-COBRAS, which integrates logical reasoning. By applying logical rules, the algorithm autonomously infers new relationships between data points from user-provided constraints, significantly reducing the need for user interaction. Furthermore, LR-COBRAS introduces a dynamic super-instance refinement process and custom data structures to adaptively adjust clustering granularity, enhancing flexibility, accuracy, and overall time efficiency. Extensive experiments on benchmarks such as CIFAR100, ImageNet, and UCI datasets demonstrate that LR-COBRAS not only reduces user interaction but also significantly improves query efficiency and time efficiency while maintaining high clustering quality. This work presents a novel approach that leverages logical reasoning to optimize interactive clustering, particularly for complex and large-scale datasets. The source code for our implementation is available on <a href="https://github.com/cjw-bbxc/LR-COBRAS">https://github.com/cjw-bbxc/LR-COBRAS</a>.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Logical reasoning-enhanced interactive clustering: an efficient algorithm for large-scale datasets

  • 期刊论文

  • 发表日期：

    2025年6月20日

查看全文
  • Ning Zhou,
  • Jiawei Cao

摘要

To address the challenges of high user involvement and low time efficiency in handling large datasets, we propose an enhanced interactive clustering algorithm called LR-COBRAS, which integrates logical reasoning. By applying logical rules, the algorithm autonomously infers new relationships between data points from user-provided constraints, significantly reducing the need for user interaction. Furthermore, LR-COBRAS introduces a dynamic super-instance refinement process and custom data structures to adaptively adjust clustering granularity, enhancing flexibility, accuracy, and overall time efficiency. Extensive experiments on benchmarks such as CIFAR100, ImageNet, and UCI datasets demonstrate that LR-COBRAS not only reduces user interaction but also significantly improves query efficiency and time efficiency while maintaining high clustering quality. This work presents a novel approach that leverages logical reasoning to optimize interactive clustering, particularly for complex and large-scale datasets. The source code for our implementation is available on https://github.com/cjw-bbxc/LR-COBRAS.