Naeini Mahdi Pakdaman, Cooper Gregory F, Hauskrecht Milos
Intelligent Systems Program, University of Pittsburgh, PA, USA.
Intelligent Systems Program, University of Pittsburgh, PA, USA ; Department of Biomedical Informatics, University of Pittsburgh, PA, USA.
Proc AAAI Conf Artif Intell. 2015 Jan;2015:2901-2907.
Learning probabilistic predictive models that are well calibrated is critical for many prediction and decision-making tasks in artificial intelligence. In this paper we present a new non-parametric calibration method called Bayesian Binning into Quantiles (BBQ) which addresses key limitations of existing calibration methods. The method post processes the output of a binary classification algorithm; thus, it can be readily combined with many existing classification algorithms. The method is computationally tractable, and empirically accurate, as evidenced by the set of experiments reported here on both real and simulated datasets.
学习经过良好校准的概率预测模型对于人工智能中的许多预测和决策任务至关重要。在本文中，我们提出了一种新的非参数校准方法，称为贝叶斯分位数装箱法（BBQ），该方法解决了现有校准方法的关键局限性。该方法对二元分类算法的输出进行后处理；因此，它可以很容易地与许多现有的分类算法相结合。该方法在计算上易于处理，并且在经验上是准确的，正如我们在此处报告的关于真实和模拟数据集的一组实验所证明的那样。
