arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2409.17643stat.MLcs.LG
收藏

高效公平性-性能帕累托前沿计算

Efficient Fairness-Performance Pareto Front Computation

Technion - Israel Institute of Technology(技术离子-以色列理工学院);NVIDIA(英伟达)

Mark Kozdoba, Binyamin Perets, Shie Mannor

AI总结:

本文提出一种无需训练复杂表示模型即可高效计算公平性-性能最优帕累托前沿的方法,利用结构性质简化为离散问题,并通过凹凸规划求解,可作为表示学习算法的基准。

摘要:

众所周知,表示的公平性与从该表示派生的分类器的性能之间存在固有的权衡。由于大多数现代表示学习方法中优化算法的复杂性,对于给定方法,判断所获得的公平性-性能曲线是否最优(即是否接近底层数据分布下这些量的真实帕累托前沿)可能并非易事。本文提出了一种计算最优帕累托前沿的新方法,该方法不需要训练复杂的表示模型。我们证明了最优公平表示具有若干有用的结构性质,这些性质使得帕累托前沿的计算可以简化为一个紧凑的离散问题。我们还进一步证明,这些紧凑的近似问题可以通过现成的凹凸规划方法高效求解。由于我们的方法独立于具体的表示模型,因此可以用作表示学习算法进行比较的基准。我们在多个真实世界基准数据集上对该方法进行了实验评估。

英文摘要:

There is a well known intrinsic trade-off between the fairness of a representation and the performance of classifiers derived from the representation. Due to the complexity of optimisation algorithms in most modern representation learning approaches, for a given method it may be non-trivial to decide whether the obtained fairness-performance curve of the method is optimal, i.e., whether it is close to the true Pareto front for these quantities for the underlying data distribution. In this paper we propose a new method to compute the optimal Pareto front, which does not require the training of complex representation models. We show that optimal fair representations possess several useful structural properties, and that these properties enable a reduction of the computation of the Pareto Front to a compact discrete problem. We then also show that these compact approximating problems can be efficiently solved via off-the shelf concave-convex programming methods. Since our approach is independent of the specific model of representations, it may be used as the benchmark to which representation learning algorithms may be compared. We experimentally evaluate the approach on a number of real world benchmark datasets.

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。

↑