Element based redescription set exploration

Functionality

The element based exploration uses the Self Organising map to order elements into different clusters based on element occurence in redescription support sets. The exploration process can be done in several steps:

1. Click on any SOM cluster to examine redescriptions describing at least one element from the cluster
2. Scroll through redescription table and select some interesting redescription
3. Examine additional information about the selected redescription
4. Create SOM map on various subests of redescriptions.

Clicking on some of the components displays new elements on the page. The elements are placed below currently displayed components, so depending on the screen size, you might need to scroll to use them.

This exploration technique is demonstrated on the DBLP dataset . This dataset contains 6455 entities (authors of scientific papers) described by two views containing Boolean attributes without missing values. The first view contains 304 descriptors describing which author published papers on which conferences. The second view contains information about co-authorships between 6455 different authors of scientific papers.

Number of rows: Number of columns: Learning rate: Number of iterations: Progress:
Train SOM on all data Train SOM on selection Save selection Load and create SOM Save SOM layout Load SOM layout

Cluster el. Redescription el. minimal Fraction: Perform