CrossWalk: Fairness-Enhanced Node Representation Learning

The potential for machine learning systems to amplify social inequities and unfairness is receiving increasing popular academic attention. Much recent work has focused on developing algorithmic tools assess mitigate such unfairness. However, there little enhancing fairness in graph algorithms. Here, we develop a simple, effective general method, CrossWalk, that enhances of various algorithms, including influence maximization, link prediction node classification, applied embeddings. CrossWalk applicable any random walk based representation algorithm, as DeepWalk Node2Vec. key idea bias walks cross group boundaries, by upweighting edges which (1) are closer the groups’ peripheries or (2) connect different groups network. pulls nodes near towards their neighbors from other embedding space, while preserving necessary structural properties graph. Extensive experiments show effectiveness our algorithm enhance classification synthetic real networks, with only very small decrease performance.