Machine Learning for Finance

Lead Research Organisation: University College London
Department Name: Computer Science

Abstract

The project will look at Gaussian Processes/ T-Processes and Deep Learning, applied within the financial arena. Currently Gaussian Processes are unable to scale beyond a few thousand training examples because of a matrix inversion. Recent research on Bayesian Deep Learning and Gaussian Processes offers both links between the two as well as the promise of scaling GPs. The Bayesian treatment of GP's offers particular promise in the financial arena due to the extra output of not only the mean function but the full probability distribution of the forecast, thus giving a confidence around that mean. The Bayesian treatment also offers the promise of training without over-fitting, together with the potential to train without cross-validation by maximising the marginal likelihood, again attractive within the financial domain.

Publications

10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/R512400/1 01/10/2017 30/09/2021
1926467 Studentship EP/R512400/1 25/09/2017 30/09/2021 John Goodacre