We propose a hierarchical framework which extracts local and global features from different scales of given graph at the same time.

Reducing the computational complexity with processing at higher levels of abstraction.

Independent processing of different resolutions of graph based on application.

Bridging the gap between elementary features and more descriptive ones.