Robert MaidstonePhD student
My project is on “Efficient Analysis for Complex Changepoint Problems” and is supervised jointly by Prof. Paul Fearnhead and Prof. Adam Letchford. The project combines statistical time series analysis methods with optimisation techniques from OR.
My PhD focuses on developing efficient methods to solve complex changepoint problems.
The idea of a changepoint is that when considering a time series changes in structure may occur. Examples of some of these changes in structure can be seen in the figure below. The places where these changes of structure occur are called changepoints and it is required for many reasons to detect how many of these occur in a time series and their locations.
Many methods for detecting these changepoints exist ranging from approximate (heuristic) fast methods, to exact methods which take longer to run. The holy grail of methods would be both fast and exact; one of the ways in which this can be achieved is to use a technique called dynamic programming.
My PhD aims to look at these methods and ways of applying them to different situations. In particular applying them to situations where the models applied to individual segments are dependent on one another.
My personal webpage: http://www.lancs.ac.uk/~maidston
On optimal multiple changepoint algorithms for large data
Maidstone, R., Hocking, T., Rigaill, G., Fearnhead, P. 5/09/2014 In: arxiv.org.