This paper considers the task of learning users' preferences on a combinatorial set of alternatives, as generally used by online configurators, for example. In many settings, only a set of selected alternatives during past interactions is available to the learner. Fargier et al. [2018] propose an approach to learn, in such a setting, a model of the users' preferences that ranks previously chosen alternatives as high as possible; and an algorithm to learn, in this setting, a particular model of preferences: lexicographic preferences trees (LP-trees). In this paper, we study complexity-theoretical problems related to this approach. We give an upper bound on the sample complexity of learning an LP-tree, which is logarithmic in the number of attributes. We also prove that computing the LP tree that minimises the empirical risk can be done in polynomial time when restricted to the class of linear LP-trees.
翻译:本文件考虑了学习用户对一组组合式替代物的偏好任务,例如在线配置器通常使用的一组替代物。在许多环境中,学习者在以往的互动中只能得到一组选定的替代物。Fargier等人([2018] )提出一种方法,在这种环境中学习一种将先前选择的替代物排在尽可能高的用户偏好模式;以及一种算法,在此背景下学习一种特定的偏好模式:词典偏好树(LP-Trees)。在本文中,我们研究了与这一方法有关的复杂理论问题。我们给学习LP-Tree的样本复杂性设定了一个上限,这是属性数的对数。我们还证明计算LP树可以将经验风险最小化到限于线性LP-Tree类的多元时间。