资源分类
》 资源 》论文》Maxing and Ranking with Few Assumptions

Maxing and Ranking with Few Assumptions

2020-02-10 | |  182 |   43 |   0

Abstract

 PAC maximum selection (maxing) and ranking of n elements via random pairwise comparisons have diverse applications and have been studied under many models and assumptions. With just one simple natural assumption: strong stochastic transitivity, we show that maxing can be performed with linearly many comparisons yet ranking requires quadratically many. With no assumptions at all, we show that for the Borda-score metric, maximum selection can be performed with linearly many comparisons and ranking can be performed with image.png comparisons.

上一篇：Non-Stationary Spectral Kernels

下一篇：Preventing Gradient Explosions in Gated Recurrent Units

用户评价
全部评价

热门资源

  • The Variational S...

    Unlike traditional images which do not offer in...

  • Learning to Predi...

    Much of model-based reinforcement learning invo...

  • Stratified Strate...

    In this paper we introduce Stratified Strategy ...

  • Learning to learn...

    The move from hand-designed features to learned...

  • A Mathematical Mo...

    Direct democracy, where each voter casts one vo...