machine learning - Representing graph as a vector -

i working on applying machine learning algorithms software fault prediction. in research, software/ software components (like classes, packages etc.) represented in form of graphs (eg. control flow graphs, data flow graphs etc.). have specific requirement convert these graphs vectors in order facilitate application of machine learning algorithms (eg. classification, clustering). understand there many graph mining algorithms need not require representing graphs vectors. but, our problem better solved if graphs represented vectors. hence, please suggest existing works/ papers this. thanks.

there no valid answer such question. graphs cannot represented finite length vectors. can extract graph features , encode them particular dimensions, features, , how represented - depends compeltely on particular task , later used method of classification/clustering. listing works doing such transformation pointless, there has been hundreads of such approaches, none of considered standard (or "valid") way so. in general encoding graphs vectors rather a bad idea. these complex structures, without varefoul, experts based analysis, cannot correctly compressed such primitive form real valued vectors. suggest, however, @ kernel based methods (there kernel methods both classification , clustering avaliable) there (reasonable) graph kernels