Calibration via Regression
Calibration via Regression
复制标题
通过回归进行校准
DOI：
10.1109/itw.2006.1633786
复制
发表时间：
2006
期刊：
影响因子：
--
通讯作者：
S. Kakade
中科院分区：
文献类型：
--
作者：
Dean Phillips Foster;S. Kakade
在在线预测设置中，具有经验（条件）频率的校准概念需要与所声称的预测概率相匹配。这与更传统的在线预测目标形成鲜明对比，即累积损失低。这些目标之间的差异通常使它们很难彼此比较。本文展示了如何从传统的在线损失最小化算法（即在线回归）中获得大概的校准形式。作为推论，我们展示了如何在子序列集合中构建校准的预测。
In the online prediction setting, the concept of calibration entails having the empirical (conditional) frequencies match the claimed predicted probabilities. This contrasts with more traditional online prediction goals of getting a low cumulative loss. The differences between these goals have typically made them hard to compare with each other. This paper shows how to get an approximate form of calibration out of a traditional online loss minimization algorithm, namely online regression. As a corollary, we show how to construct calibrated forecasts on a collection of subsequences.