arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2405.16752cs.LGcs.AI
收藏

约束优化的模型集成

Model Ensembling for Constrained Optimization

University of Pennsylvania(宾夕法尼亚大学)

Ira Globus-Harris, Varun Gupta, Michael Kearns, Aaron Roth

AI总结:

针对多维预测用于下游约束优化的场景,提出基于多校准的两种模型集成算法(白盒与黑盒),保证收敛与效用,且优于单个最佳模型。

摘要:

在机器学习中,模型集成有着悠久的历史,从 boosting 和 bagging 开始,一直延续至今。这段历史中的大部分工作都集中在组合用于分类和回归的模型上,但最近人们对更复杂的设置产生了兴趣,例如在强化学习中集成策略。同时,集成与多校准技术之间也出现了紧密的联系。在这项工作中,我们通过考虑一种设置来进一步研究这些主题,在该设置中,我们希望集成用于多维输出预测的模型,而这些预测又用于下游优化。更精确地说,我们假设给定若干模型,它们将状态空间映射到多维实值预测。这些预测构成了一个线性目标函数的系数,我们希望在该目标函数下,在指定约束条件下进行优化。我们解决的基本问题是如何改进和组合这些模型,使其在下游优化问题中的表现优于其中最好的模型。我们应用多校准技术,得出了两种可证明高效且收敛的算法。第一种(白盒方法)要求给定将状态映射到输出预测的模型,而第二种(黑盒方法)仅要求策略(从状态到优化问题解的映射)。对于这两种方法,我们都提供了收敛性和效用保证。最后,我们在受控实验环境中研究了这两种算法的性能和表现。

英文摘要:

There is a long history in machine learning of model ensembling, beginning with boosting and bagging and continuing to the present day. Much of this history has focused on combining models for classification and regression, but recently there is interest in more complex settings such as ensembling policies in reinforcement learning. Strong connections have also emerged between ensembling and multicalibration techniques. In this work, we further investigate these themes by considering a setting in which we wish to ensemble models for multidimensional output predictions that are in turn used for downstream optimization. More precisely, we imagine we are given a number of models mapping a state space to multidimensional real-valued predictions. These predictions form the coefficients of a linear objective that we would like to optimize under specified constraints. The fundamental question we address is how to improve and combine such models in a way that outperforms the best of them in the downstream optimization problem. We apply multicalibration techniques that lead to two provably efficient and convergent algorithms. The first of these (the white box approach) requires being given models that map states to output predictions, while the second (the \emph{black box} approach) requires only policies (mappings from states to solutions to the optimization problem). For both, we provide convergence and utility guarantees. We conclude by investigating the performance and behavior of the two algorithms in a controlled experimental setting.

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。

↑