I am interested in solving the following problem using a learning to hash algorithm:
I've got several attibuted graphs (a graph with node and edge labels - you can think about that as a description logic query) and I want to perform similarity search between a query and a document (both represented as an attributed graph).
After googling a lot, I found only a Kernelized Locality-Sensitive Hashing algorithm (KLSH) solution, the following one: Kernelized Hashcode Representations for Relation Extraction.
Locality-Sensitive Hashing demands more bits than a learning algorithm to represent data. It's not an ideal solution.
Is there a learning to hash algorithm to perform similarity search for description logic queries?