Combining near-optimal feature selection with gSpan

Graph classification is an increasingly important step in numerous application domains, such as function prediction of molecules and proteins, computerized scene analysis, and anomaly detection in pro