Hub Miner 1.1

Nenad Tomasev — Thu, 22 Jan 2015 16:33:51 -0000

Hub Miner is an OpenML-compatible machine learning library with an emphasis on instance-based high-dimensional data analysis under the asymmetric distribution of relevance, that is referred to as hubness.

Some updates to the previous version: - BibTex support for all algorithm implementations, making all of them easy to reference (via algref package).

Two more hubness-aware approaches (meta-metric-learning and feature construction)
An implementation of Hit-Miss networks for analysis.
Several minor bug fixes.
The following instance selection methods were added: HMScore, Carving, Iterative Case Filtering, ENRBF.
The following clustering quality indexes were added: Folkes-Mallows, Calinski-Harabasz, PBM, G+, Tau, Point-Biserial, Hubert's statistic, McClain-Rao, C-root-k.
Some more experimental scripts have been included.
Extensions in the estimation of hubness risk.
Alias and weighted reservoir methods for weight-proportional random selection.

OpenML compatibility assures that the classification experiments can be performed in a networked manner and the results can be uploaded to OpenML servers for cross-comparisons with other systems and implementations. This same functionality has recently been added to Weka, R and RapidMiner.

The library implements a series of recently proposed hubness-aware methods for learning in many dimensions.

Hub Miner implements custom methods for classification, clustering, metric learning, instance selection, outlier detection and other common machine learning tasks, as well as a set of baselines and a powerful experimental framework that includes testing under various challenging conditions.

It is implemented entirely in Java, so it should be easily portable to all platforms.

Hub Miner also contains Image Hub Explorer, a tool for experimentation with quantized image feature representations that has been presented at the European Conference on Machine Learning in 2013.

Hub Miner contains all the methods that I have proposed in my Ph.D. thesis while studying at the Artificial Intelligence Laboratory, but it actually contains a lot more than that.

There is a detailed user manual at the GitHub repo, so all new users will be able to find their way in the library with relative ease.

This is an announcement of the first release.

Future releases are expected to contain more standard baselines and more hubness-aware methods, as well as better support for handling large-scale datasets.

For more details, You are also invited to visit the Hub Miner page on my homepage: http://ailab.ijs.si/nenad_tomasev/hub-miner-library/

Feel free to contact me with comments, suggestions and ideas, as well as queries on how to use the software if You are interested.

mloss.org Hub Miner

Hub Miner 1.1