Consistent and fast inference in compartmental models of epidemics using Poisson Approximate Likelihoods

传染病动态的量化和特征化对于向官方决策者通报应对新出现的流行病至关重要；它们对于了解以前的疫情以更好地为未来做好准备也至关重要。对疾病在人群中的传播进行建模的最流行的范例是区隔模型。除了最简单的情况外，这些模型的可能性在所有情况下都是不可接近的，因此，为了进行推理，需要进行近似。在过去的几十年里，计算的进步导致了许多复杂而昂贵的模拟算法的发展，用于对随机区室模型进行推理。这个项目的目的是提出一类计算上便宜和简单的替代这些方法，这些方法被严格的一致性理论证明是合理的，并且在现实世界的数据示例中被证明是实用的。

The quantification and characterisation of infectious disease dynamics are essential for informing official decision makers in their response to emerging epidemics; they are also crucial in understanding previous outbreaks to better prepare for the future. The most popular paradigm for modelling the spread of a disease through a population is that of compartmental models. The likelihood for such models is inaccessible in all but the simplest cases, therefore, in order to perform inference, one needs to make approximations. Over the past few decades, computational advances have led to the development of many sophisticated and expensive simulation algorithms for inference on stochastic compartmental models. The aims of this project are to propose a class of computationally cheap and simple alternatives to these methods which are justified by rigorous consistency theory and demonstrated to be practically useful on real world data examples.