可混性的几何学
The Geometry of Mixability
摘要:
可混合的损失函数在在线环境中利用专家建议进行预测的背景下具有根本的重要性,因为它们表征快速学习率。通过从微分几何的角度重新解释适当性,我们提供了二元和多类情况下混合性的简单几何表征:适当的损失函数$\ell$是$η$可混合的当且仅当缩放损失函数$η\ell$的超预测集$\textrm{spr}(η\ell)$在对数损失$\ell_{\log}$的超预测集$\textrm{spr}(\ell_{\log})$内自由滑动,在关于$\ell$的可微性的相当一般的假设下。我们的方法提供了一种方法来处理一些概念,损失函数(如properness)在“坐标自由”的方式和调和以前的结果得到的混合损失函数的二进制和多类的情况下。
英文摘要:
Mixable loss functions are of fundamental importance in the context of prediction with expert advice in the online setting since they characterize fast learning rates. By re-interpreting properness from the point of view of differential geometry, we provide a simple geometric characterization of mixability for the binary and multi-class cases: a proper loss function $\ell$ is $η$-mixable if and only if the superpredition set $\textrm{spr}(η\ell)$ of the scaled loss function $η\ell$ slides freely inside the superprediction set $\textrm{spr}(\ell_{\log})$ of the log loss $\ell_{\log}$, under fairly general assumptions on the differentiability of $\ell$. Our approach provides a way to treat some concepts concerning loss functions (like properness) in a ''coordinate-free'' manner and reconciles previous results obtained for mixable loss functions for the binary and the multi-class cases.
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。