This website requires JavaScript.

Learn to Explore: on Bootstrapping Interactive Data Exploration with Meta-learning

Yukun CaoXike XieKexin Huang
Dec 2022
摘要
Interactive data exploration (IDE) is an effective way of comprehending bigdata, whose volume and complexity are beyond human abilities. The main goal ofIDE is to discover user interest regions from a database through multi-roundsof user labelling. Existing IDEs adopt active-learning framework, where usersiteratively discriminate or label the interestingness of selected tuples. Theprocess of data exploration can be viewed as the process of training aclassifier, which determines whether a database tuple is interesting to a user.An efficient exploration thus takes very few iterations of user labelling toreach the data region of interest. In this work, we consider the dataexploration as the process of few-shot learning, where the classifier islearned with only a few training examples, or exploration iterations. To thisend, we propose a learning-to-explore framework, based on meta-learning, whichlearns how to learn a classifier with automatically generated meta-tasks, sothat the exploration process can be much shortened. Extensive experiments onreal datasets show that our proposal outperforms existing explore-by-examplesolutions in terms of accuracy and efficiency.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答