Initializing Node Types

In the modifying configuration section, we have utilized ‘random with weight’ initialization method of Crowd. In this section, we explain the details of all 4 initalization methods.

With all initialization types, the total node count should be equal to network’s node count, or if setting with weights, to 1.

1. Random with count

For this initialization type, the exact number of nodes for each type is entered by the modeler. The node type will be chosen randomly for each node.

nodetypes:
    Susceptible:
        random-with-count:
            count: 90
    Infected:
        random-with-count:
            count: 10
    Recovered:
        random-with-count:
            count: 0

2. Random with weight

For this initialization type, a weight for each type is entered by the modeler. Node count for each type will be calculated by (node_count * weight_of_that_type).

The node type will be chosen randomly for each node.

nodetypes:
    Susceptible:
        random-with-weight:
            initial-weight: 0.9
    Infected:
        random-with-weight:
            initial-weight: 0.1
    Recovered:
        random-with-weight:
            initial-weight: 0

3. Choose with metric

Various social network analysis (SNA) metrics can also be utilized to choose node types.

Metric	Conf. File Notation
PageRank	pagerank
Degree centrality	degree
Betweenness centrality	betweenness
Closeness centrality	closeness
Eigenvector centrality	eigenvector
Katz centrality	katz

With these metrics, top k (count) nodes can be selected as the spreader. In the influence maximization simulations where researchers wish to compare their novel algorithms with the existing methods, centrality metrics are often chosen as a baseline.

With Crowd, researchers can simply rerun their simulations with the same settings by only changing the initialization method in the configuration file. Reading from file option, explained in the following subsection, can be used to load the seed nodes determined with the researchers’ own algorithm, or other methods to compare that are not provided by Crowd.

An example configuration is given below:

nodetypes:
    Active:
        choose-with-metric:
            metric: pagerank
            count: 10
    Inactive:
        random-with-count:
            count: 90

4. Load from file

With this option, modelers can utilize the node lists saved previously to the project’s datasets. This file can simply be copied to the project’s datasets folder or uploaded through the user interface.

Important: The file should only include the integers representing node IDs. Each number should be separated with comma or new line. Only CSV file format is accepted.

nodetypes:
    Active:
        from-file:
            path: 'name_of_file.csv'
    Inactive:
        random-with-count:
        count: 90

Next: (Optional) Crowd’s network models