Probabilistic Scientific Computing: Statistical inference approaches to numerical analysis and algorithm design
(June 5 – 9, 2017)
There is an urgent and unmet need to formally analyze, design, develop and deploy advanced methods and algorithms that can scale in statistical and computational efficiency to the size of modern data sets and the complexity of contemporary mathematical models. Addressing this need will require a holistic approach involving new foundational theory, algorithms, and programming language design.
The emerging research theme of Probabilistic Scientific Computing (PSC) or Probabilistic Numerics lies at the nexus of these overlapping directions. It aims to improve statistical quantification of uncertainty, improve computational efficiency, and build more effective and scalable numerical methods for statistical models by leveraging the natural correspondence between computation and inference.
The primary goal of the workshop is to introduce recent results and new directions in probabilistic scientific computing to the US research communities in statistics and machine learning, in numerical analysis, and in theoretical computer science.