Any views expressed within media held on this service are those of the contributors, should not be taken as approved or endorsed by the University, and do not necessarily reflect the views of the University in respect of any particular issue.

Todd Waugh Ambridge: Global Optimisation via Constructive Real Numbers

Efficient gradient descent-style local optimisation algorithms have been studied extensively and applied abundantly throughout the field of deep learning. On the other hand, convergent global optimisation algorithms that never yield an incorrect result — such as branch-and-bound methods — are much less practically available. Optimisation is a core component of supervised machine learning, having broad applications to function approximation algorithms such as (non-parametric) interpolation and (parametric) regression. Convergence properties of interpolation have been studied extensively by Weierstrass-style theorems, though the convergence of regression has not been widely studied.

In this talk, I will introduce the concepts behind convergent global optimisation and discuss how floating-point real numbers are an inappropriate data type for such algorithmics. This leads us towards exploring data types for arbitrary-precision constructive real numbers, and showing how convergent global optimisation algorithms can be built using these types. Lastly, I will apply this framework to supervised machine learning, and show how convergence properties for regression on arbitrary ‘searchable’ data types can be satisfied via convergent global optimisation.

Date:  2021/03/11

SlidesConvergent Global Optimisation via Constructive Real Numbers



Report this page

To report inappropriate content on this page, please use the form below. Upon receiving your report, we will be in touch as per the Take Down Policy of the service.

Please note that personal data collected through this form is used and stored for the purposes of processing this report and communication with you.

If you are unable to report a concern about content via this form please contact the Service Owner.

Please enter an email address you wish to be contacted on. Please describe the unacceptable content in sufficient detail to allow us to locate it, and why you consider it to be unacceptable.
By submitting this report, you accept that it is accurate and that fraudulent or nuisance complaints may result in action by the University.