Convert a statitical ranking algorithm from Matlab To Java

A recent new preference ranking algorithm has been proposed by authors Volkovs and Zemel(see link below). The algorithm is quite statistically complex and detailed and the aim of this job is to convert the authors' published Matlab code into a Java implementation.

As a minimum you should be able to read and understand (at a general level) their algorithm defined at:

http://www.cs.toronto.edu/~mvolkovs/www2012_rankaggr.pdf

This means understanding the maximum likelihood models as well as the prior models. In order to be considered for the bid you should have at least 5 years practical Java (J2SE) and Matlab experience to convert the code(which will be available if the bid goes ahead). Anyone without this is likely to struggle with the conversion. This job is ideal for a PhD or postdoc with a computer science /stats background.

The Java implementation shall match the Matlab code exactly in terms of functionality but not necessarily performance. i.e every function should produce the same results in both languages using the provided test data (give or take sampling errors used by the algorithm) It is also expected that the Java code is well written out with unit tests and commented using Javadoc.

You are not required to improve the performance of the algorithm nor implement a multithreaded version nor use J2EE. EJML is preffered as a matrix library but others are acceptable. Code, documentation, tests and test data produced as a result of any bids accepted will remain property and copyright of this bid poster i.e. me. An estimated time for this project is about 2 to 3 days for a fluent matlab and Java programmer but the bid is for a fixed amount. There are approx 800 lines of code.