Graph-Based Relational Learning with a Polynomial Time Projection Algorithm

Abstract

The paper presents a new projection operator for graphs, named AC-projection, which exhibits good complexity properties as opposed to the graph isomorphism (Θ-subsumption) operator typically used in graph mining. We study the size of the search space and some practical properties of the projection operator. These properties give us a specialization algorithm using simple local operations. Then we prove experimentally that we can achieve an important performance gain (polynomial complexity projection) without or with non-significant loss of discovered patterns quality.

LC-mine: a framework for frequent subgraph mining with local consistency techniques

“…The first instance, named FGMAC [4], follows a breadth-first order to find frequent subgraphs and uses an Apriori-like [1] search strategy. The second, named AC-miner [5], is a patterngrowth approach that follows a depth-first search space exploration strategy and uses powerful pruning techniques in order to considerably reduce this search space. These two instances are based on the arc consistency technique.…”

“…In this section, we will present, AC-miner [5], a basic pattern-growth instance of the LC-mine framework for frequent AC-reduced subgraphs mining. This approach is based on three key concepts: the AC-extension operator (member of the LC-mine framework), the increasing forbidden labels inheritance and the decreasing allowed vertices inheritance.…”

“…The second is a pattern-growth approach that follows a depth-first search space exploration strategy and uses powerful pruning techniques in order to considerably reduce this search space. This paper is a wide extension of two previous works we carried out [4,5]. The present work defines a generic framework that generalizes the two instances presented in [4] and [5].…”

“…This paper is a wide extension of two previous works we carried out [4,5]. The present work defines a generic framework that generalizes the two instances presented in [4] and [5]. We have specified this framework after a thorough analysis of the characteristics of the two approaches.…”

“…We have specified this framework after a thorough analysis of the characteristics of the two approaches. Furthermore, the earlier works have been extended in two ways: First, we have brought theoretical proofs related to the pruning techniques initially introduced in [5]. Then, we have performed an extensive comparative study of the two proposed approaches with much larger graph databases.…”