GREENCYCLESII
Vacancies in related projects
3-month internship - Inferring the earth system, now and into the future.
Microsoft Research, Cambridge, UK
Society urgently requires accurate predictions of how the living systems of the earth are going to interact-with and influence climate change. Unfortunately earth system models have not been built in a formal probabilistic framework, making it unclear how much confidence to place in their predictions. Our Computational Ecology group is working on constructing models of the living systems of the earth while accurately quantifying uncertainty in the source data, the models, and their predictions. The goal is to enable more accurate predictions by providing a framework for the earth system modelling community to rapidly develop and simulate models and robustly quantify their predictive accuracy. The adoption of accessible computational methods for the required Bayesian inference will be an important part of the success of this project. Our Machine Learning and Perception group is working on such general frameworks for Bayesian inference, one of which is the Infer.NET framework. However, to date, the potential benefits to the earth-system modelling community of adopting Infer.NET as a framework has not been explored. In this internship you will work with both the Computational Ecology and Machine Learning and Perception groups to explore the extent to which their research into improving earth system models might benefit from using Infer.NET for Bayesian inference. The project will have two main tasks i) To construct a graphical model of a simplified earth system model which the Computational Ecology group has already developed and ii) to use Infer.NET to perform Bayesian inference on that model. The intended outcomes will be one or two scientific papers reporting the findings of this research.
This is a 3 month paid internship at Microsoft Research Cambridge, UK and would ideally suit PhD/Masters level students with a statistics background. Please contact Matthew.Smith@microsoft.com for more details.