Today, I managed to push out a range of things…
For one thing, I finally sat down and cleaned up and consolidated the documentation of my python-weka-wrapper3 library. The instructions are now always aimed at creating a virtual environment rather than installing the library system-wide (and removed the page on virtual environments). Using virtual environments simply avoids the problem of accidentally stuffing up your system libraries (been there, done that and it’s not pretty!). Code examples were sort of duplicated between the API and Examples pages, so I merged them into just the Examples page. The Source code page has been obsolete for a long time, so I retired it as well. A lot happier with it now. 🙂
With a new release of OpenSSH being available (to fix a high-severity flaw), I updated my Windows 32/64bit binaries of my rsync4j library. But I didn’t stop there! Something that always irked me was that the get/set methods in the Java class wrappers for the binaries didn’t actually mention in the Javadoc what command-line options of the actual binary they were corresponding with. Well, that was boring and took quite a while (ssh, ssh-key-gen and rsync have gazillions of options!), but it’s finally done! The newest release is now: 3.2.3-3
And finally something new: a couple of weeks ago, I embarked on an experiment, whether I could integrate Weka within scikit-learn. Since my python-weka-wrapper3 project already made Weka available in Python-land, I thought that it shouldn’t be too hard. And it wasn’t! 😉 It just took a while to get used to the scikit-learn API and how estimators etc have to be written in order to work with cloning etc. The first release of the sklearn-weka-plugin offers the following functionality: classification, regression, clustering, filtering. Let’s see how that library evolves!