arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2403.04015cs.LGcs.AIstat.ML
收藏

通过单个预训练强化智能体实现的Knockoff引导特征选择

Knockoff-Guided Feature Selection via A Single Pre-trained Reinforced Agent

School of Computing and Augmented Intelligence, Arizona State University(亚利桑那州立大学计算与增强智能学院);Department of Computer Science, University of Kansas(堪萨斯大学计算机科学系);Department of Statistics and Data Science, University of Central Florida(中佛罗里达大学统计与数据科学系);NEC Laboratories America Inc(美国电气公司实验室)

Xinyuan Wang, Dongjie Wang, Wangyang Ying, Rui Xie, Haifeng Chen, Yanjie Fu

AI总结:

针对现有有监督和无监督特征选择方法的缺陷,提出由knockoff特征引导、单个预训练强化智能体优化的特征选择框架,通过伪标签从三方面改进选择过程,提升效率并降低对目标变量的依赖。

摘要:

特征选择通过消除冗余特征来保障数据的AI就绪度。现有研究主要分为两类:i)有监督特征选择(Supervised Feature Selection, SFS),其基于特征与目标变量的相关性来识别最优特征子集;ii)无监督特征选择(Unsupervised Feature Selection, UFS),其通过捕捉特征集内的核心信息而非利用目标变量来降低特征空间维度。然而,SFS方法因依赖目标变量和下游机器学习任务,存在过程耗时且泛化性有限的问题;UFS方法则受限于其得到的降维特征空间是隐式且不可追溯的。为应对这些挑战,我们提出了一种创新的特征选择框架,该框架由knockoff(仿冒)特征引导,并通过强化学习优化,以识别最优且有效的特征子集。具体而言,我们的方法会生成“knockoff”特征,这类特征复制了原始特征的分布与特性,但与目标变量相互独立。随后,每个特征会根据其与所有knockoff特征的相关性被分配一个伪标签,作为特征评估的全新指标。我们的方法利用这些伪标签,通过三种新颖的方式引导特征选择过程,并由单个强化智能体进行优化:1)采用深度Q网络(Deep Q-network),利用原始特征及其对应的伪标签进行预训练,以提升特征选择中探索过程的效率;2)引入无监督奖励,基于伪标签和特征空间重构损失来评估特征子集质量,以降低对目标变量的依赖;3)采用一种新的ε-greedy策略,融合伪标签带来的洞见,使特征选择过程更高效。

英文摘要:

Feature selection prepares the AI-readiness of data by eliminating redundant features. Prior research falls into two primary categories: i) Supervised Feature Selection, which identifies the optimal feature subset based on their relevance to the target variable; ii) Unsupervised Feature Selection, which reduces the feature space dimensionality by capturing the essential information within the feature set instead of using target variable. However, SFS approaches suffer from time-consuming processes and limited generalizability due to the dependence on the target variable and downstream ML tasks. UFS methods are constrained by the deducted feature space is latent and untraceable. To address these challenges, we introduce an innovative framework for feature selection, which is guided by knockoff features and optimized through reinforcement learning, to identify the optimal and effective feature subset. In detail, our method involves generating "knockoff" features that replicate the distribution and characteristics of the original features but are independent of the target variable. Each feature is then assigned a pseudo label based on its correlation with all the knockoff features, serving as a novel metric for feature evaluation. Our approach utilizes these pseudo labels to guide the feature selection process in 3 novel ways, optimized by a single reinforced agent: 1). A deep Q-network, pre-trained with the original features and their corresponding pseudo labels, is employed to improve the efficacy of the exploration process in feature selection. 2). We introduce unsupervised rewards to evaluate the feature subset quality based on the pseudo labels and the feature space reconstruction loss to reduce dependencies on the target variable. 3). A new ε-greedy strategy is used, incorporating insights from the pseudo labels to make the feature selection process more effective.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑