Thomas Abeel

We have released version 0.0.10 of the Java Machine Learning Library.

New:

* two filters
* library core restructuring
* tweaks to some clustering algorithms
* included the JAMA package

This release features two new Filters. First, the Remove filter that allows you to remove certain attributes from all the instances in a Dataset. And second, the Principle Components Analysis filter, which allows you to select only the most relevant attributes of a dataset.

We have restructured the core of the library to use double arrays instead of float arrays for the default instance representation. This makes the use of the library easier.

Apart from these two new filter algorithms and the restructuring of the core, we also made some minor tweaks to other algorithms. AntBasedClustering can now determine the number of iterations based on the dataset size. SOM can now use different distance measures. The MathUtils class now takes the epsilon parameter in account when comparing two arrays of doubles. We have rewritten the source of the Manhattan distance to be more efficient.

Finally, we have also included the JAMA package in the library for the implementation of the Principle Components Analysis algorithm. Note, however, that the JAMA package is distributed as public domain software.