Variational Inference for Intractable and Misspecified State Space Models

状态空间模型（State Space Model，SSM）被广泛用于对通货膨胀和金融波动等经济变量进行建模。变分推理是一种允许快速实现SSM的技术，但其属性尚未被理解。本计画旨在研究经济学中SSMs变分推论的性质。 这项研究将开发新的变分推理技术，以提高推理和预测精度的SSM。这一项目的预期影响是，它将扩大经济机构使用更大的SSM的能力，这将使经济变量的模型更加准确。这将带来重大的社会效益，使经济政策更加知情。

State space models (SSMs) are popularly used to model economic variables such as inflation and financial volatility. Variational inference is a technique that allows for fast implementation of SSMs, but whose properties are yet to be understood. This project aims to study the properties of variational inference for SSMs used in economics. This research will develop new variational inference techniques to improve inferential and predictive accuracy from SSMs. An expected implication of this project is that it will expand the ability of economic institutions to employ larger SSMs, which will allow for more accurate models for economic variables. This will provide significant social benefits by leading to better informed economic policy.