了解你未流动的内容
Know What You Don't Flow
AI总结：
针对LHC物理生成式神经网络的不确定性校准需求，研究异方差与贝叶斯正态流在玩具模型及顶夸克对事例中对相空间密度不确定性的学习与传播方法。
AI中文摘要：
在大型强子对撞机（LHC）物理中，生成式神经网络也需要经过校准的学习不确定性。针对具有显式似然的玩具模型，我们展示了异方差正态流（heteroscedastic normalizing flow）和贝叶斯正态流如何学习基础相空间密度上的系统不确定性和统计不确定性。在无显式似然的情况下，我们针对分类器加权的近似生成式网络训练异方差损失。我们以顶夸克对事例为例说明这一综合方法，并展示条件异方差流如何将校准后的不确定性传播到所有相空间方向。
英文摘要：
Calibrated learned uncertainties are a key requirement also for generative neural networks in LHC physics. For a toy model with an explicit likelihood we show how a heteroscedastic and a Bayesian normalizing flow learn the systematic and statistical uncertainties on the underlying phase space density. Without an explicit likelihood we train the heteroscedastic loss on a classifier-reweighted approximate generative network. We illustrate our comprehensive approach for top pair events and show how a conditional heteroscedastic flow propagates calibrated uncertainties to all phase space directions.
