This website requires JavaScript.

Toward Efficient Automated Feature Engineering

Kafeng WangPengyang WangChengzhong xu
Dec 2022
摘要
Automated Feature Engineering (AFE) refers to automatically generate andselect optimal feature sets for downstream tasks, which has achieved greatsuccess in real-world applications. Current AFE methods mainly focus onimproving the effectiveness of the produced features, but ignoring thelow-efficiency issue for large-scale deployment. Therefore, in this work, wepropose a generic framework to improve the efficiency of AFE. Specifically, weconstruct the AFE pipeline based on reinforcement learning setting, where eachfeature is assigned an agent to perform feature transformation \com{and}selection, and the evaluation score of the produced features in downstreamtasks serve as the reward to update the policy. We improve the efficiency ofAFE in two perspectives. On the one hand, we develop a Feature Pre-Evaluation(FPE) Model to reduce the sample size and feature size that are two mainfactors on undermining the efficiency of feature evaluation. On the other hand,we devise a two-stage policy training strategy by running FPE on thepre-evaluation task as the initialization of the policy to avoid trainingpolicy from scratch. We conduct comprehensive experiments on 36 datasets interms of both classification and regression tasks. The results show $2.9\%$higher performance in average and 2x higher computational efficiency comparingto state-of-the-art AFE methods.
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