Mining gold from implicit models to improve likelihood-free inference
Mining gold from implicit models to improve likelihood-free inference
复制标题
DOI：
10.1073/pnas.1915980117
复制
发表时间：
2020-03-10
影响因子：
11.1
通讯作者：
Cranmer, Kyle
中科院分区：
文献类型：
--
作者：
Brehmer, Johann;Louppe, Gilles;Cranmer, Kyle
模拟器通常提供对真实世界现象的最佳描述。然而，它们隐式定义的概率密度通常是难以处理的，导致推理的逆问题具有挑战性。最近，一些技术已被引入，其中一个代理的棘手的密度是学习，包括规范化的流量和密度比估计。我们发现，表征潜在过程的额外信息通常可以从模拟器中提取，并用于增强这些代理模型的训练数据。我们引入了几个损失函数，利用这些增强的数据，并证明这些技术可以提高样本的效率和质量的推断。
Simulators often provide the best description of real-world phenomena. However, the probability density that they implicitly define is often intractable, leading to challenging inverse problems for inference. Recently, a number of techniques have been introduced in which a surrogate for the intractable density is learned, including normalizing flows and density ratio estimators. We show that additional information that characterizes the latent process can often be extracted from simulators and used to augment the training data for these surrogate models. We introduce several loss functions that leverage these augmented data and demonstrate that these techniques can improve sample efficiency and quality of inference.