scikit-learn/scikit-learn

coclust package has much better results on benchmark datasets

Open

#8,884 创建于 2017年5月15日

在 GitHub 查看
 (1 评论) (0 反应) (0 负责人)Python (66,084 star) (27,020 fork)batch import
Needs BenchmarksPerformancehelp wanted

描述

It seems that the coclust package works better than our algorithms on some benchmark datasets. We should investigate this: http://coclust.readthedocs.io/en/v0.2.1/examples.html

(I'm not sure if the benchmarks are on the website)

贡献者指南

技术栈
python
领域
datamachine learning
议题类型
research
难度面向新贡献者的预计实现难度，1 表示很小改动，5 表示专家级工作。
3
预计时间有经验贡献者完成调查、实现、测试并准备 pull request 的粗略时间范围。
1-2 days
活动状态议题当前的可参与程度：新鲜、活跃、陈旧、阻塞或等待维护者输入。
stale
清晰度议题是否清楚说明期望改动、验收标准和下一步。
needs investigation
前置要求
Pythonscikit learncoclust
新手友好度1-100 的估计分数，表示该议题对首次贡献者的友好程度。
20
研究方向
The issue asks to investigate why the coclust package yields better results on benchmark datasets compared to scikit learn's clustering algorithms. Start by reviewing the coclust documentation and examples at http://coclust.readthedocs.io/en/v0.2.1/examples.html. Identify which benchmark datasets and metrics are used. Compare the algorithms and parameters; note any differences in preprocessing, initialization, or evaluation. Look for existing scikit learn benchmarks or issues mentioning coclust. Propose a plan for reproducing the results and potentially improving scikit learn's clustering implementations. No linked PRs or comments provide further direction.