Pattern Recognition ( IF 8 ) Pub Date : 2021-07-15 , DOI: 10.1016/j.patcog.2021.108175 Alexander Korotin 1 , Vladimir V’yugin 1, 2 , Evgeny Burnaev 1
In this paper we extend the setting of the online prediction with expert advice to function-valued forecasts. At each step of the online game several experts predict a function, and the learner has to efficiently aggregate these functional forecasts into a single forecast. We adapt basic mixable (and exponentially concave) loss functions to compare functional predictions and prove that these adaptations are also mixable (exp-concave). We call this phenomenon mixability (exp-concavity) of integral loss functions. As an application of our main result, we prove that various loss functions used for probabilistic forecasting are mixable (exp-concave). The considered losses include Sliced Continuous Ranked Probability Score, Energy-Based Distance, Optimal Transport Costs & Sliced Wasserstein-2 distance, Beta-2 & Kullback-Leibler divergences, Characteristic function and Maximum Mean Discrepancies.
中文翻译:
积分损失的可混合性:有效在线聚合功能和概率预测的关键
在本文中,我们将具有专家建议的在线预测设置扩展到函数值预测。在网络游戏的每一步,都有几个专家预测一个函数,学习者必须有效地将这些函数预测聚合成一个单一的预测。我们采用基本的可混合(和指数凹)损失函数来比较功能预测,并证明这些适应也是可混合的(exp-concave)。我们称这种现象为积分损失函数的混合性(exp-concavity)。作为我们主要结果的应用,我们证明了用于概率预测的各种损失函数是可混合的(exp-concave)。考虑的损失包括切片连续排名概率分数、基于能量的距离、最佳运输成本和切片 Wasserstein-2 距离、Beta-2 和 Kullback-Leibler 分歧,