Previous Seminars


Google's Tensorflow: Not Just for Deep Learning?

More than 65 attendees from across the University came along to the seminar entitled 'Google's TensorFlow: not just for Deep Learning?' by the University of Lancaster's Dr James Hensman.​In this talk, James introduced Google's TensorFlow framework, and described how it can be used to build…

Learn More

The Zig-Zag process & Super-efficient Sampling For Bayesian Analysis Of Big Data

Standard MCMC methods can scale poorly to big data settings due to the need to evaluate the likelihood at each iteration. There have been a number of approximate MCMC algorithms that use sub-sampling

Learn More

Density Estimation: Views From Machine Learning And Statistics

Machine learning methods are getting rapidly better at fantasizing novel complex objects, such as images or audio, based on large training sets of examples. Some of the methods behind these impressive demonstrations are also useful for scientific data analysis in two ways. First, the probabilistic…

Learn More

A whirlwind tour of recent advances in changepoints

It is increasingly recognised that modern time series are not stationary. A simple departure from a stationarity assumption is a piecewise stationarity...

Learn More

Sub-quadratic search for significant correlations

Randomized algorithms have been very helpful for data reduction, with successful applications in log analysis and dimensionality reduction for machine learning. These techniques have more recently...

Learn More