In this video Iain uses animation to describes his research on developing algorithms that get computers to answer questions about data. These algorithms can sample alternative explanations, and report uncertainty in their answers.
Iain Murray develops algorithms that get computers to answer questions about data. These algorithms can sample alternative explanations, and report uncertainty in their answers. The video gives two examples of data that can be ambiguous.
The star example relates to a paper “Dynamical inference from a kinematic snapshot: the force law in the solar system”, a method for inferring unseen mass distributions from a snapshot of stellar observations (demonstrated on our sun with observations of the planets).
The chess example relates to both “Incorporating Side Information into Probabilistic Matrix Factorization Using Gaussian Processes”, a paper where basketball scores were modeled, and a Master’s project on rating systems supervised by Iain.
Much of Iain’s time has been spent developing algorithms to sample from different explanations of data. Both of the applications above have made use of one such method, “Elliptical Slice Sampling”.
Learning how to represent data in different applications is also important. Iain also has work on evaluating models of English Text and Images, and new models for representing rich, high-dimensional data. He currently has a PhD student working on machine learning of representations of acoustic data for speech recognition applications.
Find out more:
This entry was posted in: