标题： 在线预测对于流式观测数据 显示英文标题

标题： Online Prediction For Streaming Observational Data

摘要： 自动收集流式观测数据已成为标准，这使得大多数传统分析技术无法应对。 不仅模型难以识别，可能根本不存在可以安全且有用假设的模型。 事实上，经常只能做出并评估预测。 这类数据的问题通常被称为{\cal{M}}-Open，并推动了新的方法和哲学。 本文将回顾针对{\cal{M}}-Open 问题类的一些最成功的近期预测方法。 技术包括使用专家建议的预测器，如 Shtarkov 解决方案，贝叶斯非参数方法，如高斯过程先验，基于哈希函数的预测器，如{\sf 计数-最小}概要，以及符合预测。 在整个过程中，从有原则的角度展示了并比较了预测器的特性。 显示英文摘要

摘要： The automated collection of streaming observational data has become standard and defies most traditional analytic techniques. It is not just that models are hard to identify, there may not be any model that can be safely and usefully assumed. Indeed, frequently it is only predictions that can be made and assessed. Problems for this kind of data are often called {\cal{M}}-Open and have motivated new approaches and philosophies. This paper will review some of the most successful recent predictive methods for the {\cal{M}}-Open problem class. Techniques include predictors using expert advice such as the Shtarkov solution, Bayesian nonparametrics such as Gaussian process priors, hash function based predictors such as the {\sf Count-Min} sketch, and conformal prediction. Throughout, the properties of the predictors are presented and compared from a principled standpoint.