Simm Gregor N, Proppe Jonny, Reiher Markus
ETH Zürich Laboratorium für Physikalische Chemie Vladimir-Prelog-Weg 2, CH-8093 Zurich.
ETH Zürich Laboratorium für Physikalische Chemie Vladimir-Prelog-Weg 2, CH-8093 Zurich;, Email:
Chimia (Aarau). 2017 Apr 26;71(4):202-208. doi: 10.2533/chimia.2017.202.
Computational models in chemistry rely on a number of approximations. The effect of such approximations on observables derived from them is often unpredictable. Therefore, it is challenging to quantify the uncertainty of a computational result, which, however, is necessary to assess the suitability of a computational model. Common performance statistics such as the mean absolute error are prone to failure as they do not distinguish the explainable (systematic) part of the errors from their unexplainable (random) part. In this paper, we discuss problems and solutions for performance assessment of computational models based on several examples from the quantum chemistry literature. For this purpose, we elucidate the different sources of uncertainty, the elimination of systematic errors, and the combination of individual uncertainty components to the uncertainty of a prediction.
化学中的计算模型依赖于许多近似方法。这些近似方法对从它们得出的可观测值的影响通常是不可预测的。因此，量化计算结果的不确定性具有挑战性，然而，这对于评估计算模型的适用性是必要的。常见的性能统计量，如平均绝对误差，容易失效，因为它们没有将误差的可解释（系统）部分与不可解释（随机）部分区分开来。在本文中，我们基于量子化学文献中的几个例子，讨论计算模型性能评估的问题和解决方案。为此，我们阐明了不确定性的不同来源、系统误差的消除以及将各个不确定性分量组合成预测的不确定性。
