Make edges

[justaddcoffee]

Code to emit positive and negatives train/test/validation edges for ML

Given a graph (from formatted node and edge TSVs), output positive edges and negative
edges for use in machine learning.

To generate positive edges: a set of test (and optionally validation) positive edges equal in number to [(1 - train_fraction) * number of edges in input graph] are randomly selected from the edges in the input graph, such that both nodes participating in the edge have a degree greater than min_degree (to avoid creating disconnected components). These edges are emitting as positive test [and optionally positive validation] nodes. These edges are then removed from the edges from the input graph and these are the training edges.

Negative edges are selected by randomly selecting pairs of nodes that are not connected by an edge in the input graph. The number of negative edges emitted is equal to the number of positive edges emitted above.

Outputs these files in [output_dir]:

    pos_train_edges.tsv - input graph with test [and validation] positive edges
                          removed
    pos_test_edges.tsv - positive edges for test
    pos_valid_edges.tsv (optional) - positive edges for validation
    neg_train.tsv - a set of negative edges equal in number to pos_train_edges.tsv
    neg_test.tsv - a set of negative edges equal in number to pos_test_edges.tsv
    neg_valid.tsv (optional) - a set of negative edges equal in number to pos_valid_edges.tsv

    pos_train_nodes.tsv - identical to input nodes tsv

[deepakunni3]

@justaddcoffee Looks good. 👍