Work through a node2vec feature safely

Use node2vec for introduction recommendations without leaking future edges into the embedding. node2vec: biased random walks create node embeddings that preserve selected network neighborhoods. The common shortcut is to train embeddings on the full graph and evaluate on links the embedding has already seen indirectly. Choose a cutoff Training graph includes introductions through March 31. April introductions are held out for evaluation. Time splits protect the model from learning the future relationships it is supposed to predict. Tune walk bias Use one setting that favors local neighborhoods and one that explores outward structural roles. The p and q bias idea…

Sign up free — one personalized lesson every day, matched to your role and goals.

Already have an account? Sign in