Emulation methods to improve inference performance for complex models

这个项目将研究使用困难的似然景观的仿真来加速推理算法。将开发方法来引入对似然函数的多变量近似，该似然函数的计算速度比闭合形式的似然函数快得多。这些方法将被应用于蛋白质荧光的复杂模型的推断。所产生的方法将被用于利用时间分辨荧光实验的数据在光物理的几个案例研究中进行参数推断和模型排序。

This project will investigate using emulation of difficult likelihood landscapes to speed up inference algorithms. Methods will be developed to introduce a multivariate approximation to a likelihood function that can be evaluated significantly faster than a closed form likelihood. The methods will be applied to inference over complex models of protein fluorescence. Resulting methods will be used for parameter inference and model ranking in several case studies in photophysics using data from time resolved fluorescence experiments.