Fig. 7
From: Molecular complex detection in protein interaction networks through reinforcement learning
Higher values favor higher densities. A The graph on the toy network shows a positive correlation between value function and density and therefore, complexes and trajectories with higher densities are favored. The density histogram for the training phase shows a higher frequency for higher density subgraphs. B For hu.MAP 1.0, the graph shows a stronger correlation between density and value function. The density histogram again shows that higher densities are more frequently observed