CLARUS

What you'll find here.

CLARUS (Interactive Explainable Platform for Graph Neural Networks) can be used to better understand, validate and improve the decision making of Graph Neural Networks (GNN).

Therefore, CLARUS allows you to visually explore and modify the graph data in order to test the resulting effects on the GNN prediction.

CLARUS has been developed for simple, undirected, homogeneous graphs which represent a biological network.

How to use this Platform.

Select Graph Data

Choose one of our pre-selected datasets to load into the UI.

Explore the Graph

Explore the network to understand its structure and identify relevant components.

Modify the Graph

Modify the graph based on the relevances or your domain knowledge by deleting / adding nodes or edges.

Retrain or Predict

Use the modified graph data for retraining or predictions on the GNN.

Select a dataset:

Download datasets as PKL

Selected patient:

Nodes displayed:

(their next neighbours will also be displayed in the graph)

Download Results

Sort nodes:

Color edges:

Color nodes:

Modify Graph:

Delete Node

Add a new Node

Delete Edge

Add a new Edge

1. Select Node for Deletion:

1. Enter Label for the new Node:

2. Enter Values for Node Attributes (optional, otherwise Value is 0):

Hint: After adding the node, its attributes values can't be changed!

1. Select Node for Edge Deletion:

2. Select second Node to disconnect:

1. Select Node for Edge Addition:

2. Select second Node to connect to:

3. Enter Values for Edge Attributes (optional, otherwise Value is 0):

Data on Edges

Hint: One node label can occur multiple times in both columns 'from' and 'to'. Use search function to view all edges of a node.