COBRAS
Library for semi-supervised clustering using pairwise constraints
Description:
Library for semi-supervised clustering using pairwise constraints. COBRAS supports three modes for constraint elicitation:
- With labeled data. in this case the pairwise relations are derived from the labels. This is mainly used to compare COBRAS experimentally to competitors.
- With interaction through the commandline. In this case the user is queried about the pairwise relations, and can answer with yes (y) and no (n) through the commandline. The indices that are shown in the queries are the row indices in the specified data matrix (starting from zero).
- With interaction through a visual user interface. If you use COBRAS-TS, the instantiation of COBRAS that is tailored to time series clustering, you can use an interactive web application that visualizes the data, queries, and intermediate clustering results.
More Information:
- Grand Challenge/Workpackage(s): GC1 - WP1, WP2
- Responsible Research Lead: Hendrik Blockeel
- Authors: Hendrik Blockeel, Toon Van Craenendonck, Jonas Soenen, Aras Yurtman, Wannes Meert
- Programming Language: Python
- Link to the website
- Link to the repository
- Categories: Time series, active learning, anomaly detection, un/semi-supervised learning
Publications:
- They can be found here
