Sofus' Downloads Page

[NetKit-SRL] [ROCTool] [Top-K Evaluations]


NetKit-SRL is now open source on sourceforge. Please go there for source code, new releases and updates.

NetKit-SRL, or NetKit for short, is an open-source Network Learning Toolkit for statistical relational learning. It is written in Java 1.5 and is described in greater detail in the JMLR paper:

Macskassy, S. A., Provost, F. (2007) Classification in Networked Data: A toolkit and a univariate case study, Journal of Machine Learning, 8(May):935-983, 2007. [pdf].
This toolkit helps download, extract and explore the wikipedia graph in a way similar to what was describe in Discovering Users' Topics of Interest on Twitter: A First Look. It provides a web-based interface where you can type in text, extract entities from it, and explore their relationship in the Wikipedia graph.
ROCTool (ROC Bounds Toolkit)
This java-based toolkit generates confidence bounds and bands on a ROC Curve using one of various methods. It is based on the methods described in Confidence Bands for ROC Curves: Methods and an Empirical Study as well as newer methods described in a journal paper under preparation.

The source code and binary is available upon request.

Top-K Evaluations
This java-based utility program generates confidence bounds and bands based on the expected performance of the random model. It also allows to test the significance of any top-k performance as well as to highlight under which values of k a ranking is significant at a given p-value. This algorithm and implementation is described in greater detail in a journal paper which has been submitted to JAIR.

The algorithm was written in Java 1.5 and is available here: