v1.02

Minor improvements over the paper version:

• A graph-based (instead of tree-based) algorithm is available for the BuildShape phase, which can produce smaller graphs. See grg construct --bs-triplet
• There is a pygrgl.dot_product Python (and corresponding C++) method to make performing dot-products in the GRG simpler. For Python, this is also significantly faster than doing a graph traversal, since it takes NumPy arrays as inputs and then performs all the calculations in the C++ library.

GRG paper version

The version that was used for the experiments described in the GRG paper.