计算机科学
推论
干涉测量
贝叶斯推理
正规化(语言学)
自举(财务)
人工智能
反问题
压缩传感
贝叶斯概率
机器学习
计算机视觉
算法
光学
物理
数学
计量经济学
数学分析
摘要
The problem of inference in optical interferometry, i.e. turning the on-sky data into meaningful astrophysics, is a difficult ill-posed problem. But in the last two decades, several exciting developments have taken place and novel algorithms have arisen; in imaging: multi-wavelength imaging, dynamical imaging, imaging on spheroids, and production of error bars on images; in model-fitting: new bootstrapping techniques and Bayesian model selection for model-fitting. Both the characterization of the data (likelihood) and of our expectation of the solution (regularization) have improved. Buzzword-sounding techniques such as Compressed Sensing, Machine Learning, ADMM, and GPU Computing are now finding practical applications. The recent algorithmic work by the Event Horizon Telescope team has also sparked interest in optical interferometry. This paper covers these topics in an attempt to predict what the future holds for inference in our field.
科研通智能强力驱动
Strongly Powered by AbleSci AI