Arc Consistency Projection: A New Generalization Relation for Graphs
Summary (2 min read)
1 Introduction
- The complexity of the computation of the generality relation between two relational descriptions, is a crucial problem.
- For conceptual graphs [1] this operation is named projection.
- The search for an homomorphism between a tree and a graph is polynomial but between general graphs, the problem is NP complete [3] .
- This final link gives an interesting algorithmic point of view since CSP community has many results which improve the resolution algorithm.
- The authors use a new generality relation named AC-projection based on the arc consistency algorithm [12] and they prove that the search space, for a relational machine learning classification problem, is a concept lattice [13] .
2 A New Projection: AC-Projection
- Any constraint satisfaction problem can be viewed as a "network" of variables and constraints.
- Among backtracking based algorithms for constraint satisfaction problems, algorithm employing constraint propagation, like forward checking and arc consistency [12] , have had the most practical impact.
- These algorithms use constraint propagation (arc consistency) during a search to prune inconsistent values from the domains of the uninstantiated variables.
2.1 AC-Projection and Arc Consistency
- In this paragraph the authors present the arc consistency using a graph notation.
- In this paragraph the authors study some important properties of the arc consistency.
2.2 AC-Projection Properties
- The authors have defined a new mapping relation between graphs.
- The authors recall the classical homomorphism definition for digraphs.
which preserves the arcs and labels, i.e such that (x,y)
- The authors have the following proposition which links the AC-projection to the Homomorphism.
- This proposition is the foundation of many CSP resolution methods.
- In their case, the size of the largest domain is the size of the largest subset of nodes with the same label.
2.3 AC-Projection Algorithm
- The authors give a simple AC-projection algorithm for digraphs (based on AC1 algorithm [12] ).
- It begins by the creation of a first rough labeling I and reduces, for each vertex x, the given lists I(x) to consistent lists using the procedure ReviseArc.
3 AC-Projection and Machine Learning
- In this paper the generalization partial order was based on homomorphism relation between digraph.
- To deal with the homomorphism complexity, the authors proposed a class of digraph with a polynomial homomorphism operation.
- This limit the generality of the description language.
- For the homomorphism relation the interpretation is less natural since two vertices can get the same image.
- The structural interpretation of the AC-projection seems unnatural.
Fig. 2. AC-projection and interpretation
- A generalization algorithm uses a generalization operator: from two graphs the authors search for the more specific graph which generalizes two graphs (least general generalization [14] ).
- The authors generalization order will use the AC-projection relation.
- First, the authors have to specify the generalisation relation between digraphs.
Definition 6 (Generalisation relation)
- This relation is only a pre-order because the antisymmetry property is not fulfilled.
- The same problem occurs in Inductive Logic Programming.
- To get rid of this problem, Plotkin [14] defined equivalence classes of clauses, and showed that there is a unique representative of each clause, which he named 'the reduced clause'.
- For this purpose, the authors define the following equivalence relation between two digraphs.
3.2 AC-Projection and Reduction
- The authors have an equivalence relation between graphs using the AC-projection.
- For this purpose, the authors define two reduction operators.
- Using these operators the authors construct an AC-equivalent digraph by removing (first operator) or merging (second operator) vertices.
Definition 9 (AC-equivalent vertices)
- In the Figure 3 the nodes with same label are, in this case, AC-equivalent.
- These two definitions give a reduction operator.
3.3 AC-Projection and Generalization Operator
- There are some pairs, (representation languages and generality relations), which have a least general generalization operator .
- For logic formula and θ-subsumption, this operator is the classical lgg (or rlgg) introduced by plotkin [14] .
- For graph and homomorphism this operator is the graph product [10] .
Definition 12 (concept, ∨, ≥) For a set of examples E, each example e ∈ E is described by a digraph d(e) ∈ D (description space).
- This proposition gives the structure of the search space (a concept semilattice).
- The partial order between the elements of the lattice is based on AC-projection, then the digraph, intension part of a concept, can be interpreted as a compact description of a very large (potentially infinite) set of trees.
- Using this example, the authors obtain a lattice which is isomorphic with the one given by Graal [10] but it is not always the case.
- This comes from the fact that, for this set of graphs, the set of included paths is enough to obtain this lattice.
5 Conclusion
- This study has attempted to merge ideas from different communities: Graph, ILP, CSP.
- The definition of a least general generalization operator.
- -All the operations are polynomial (equivalence, reduction, product and concept order).
- -We find graphs which express a large set of trees in a compact form.the authors.the authors.
- But, since the AC projection is less precise, the classifications obtained are also less precise.
Did you find this useful? Give us your feedback
Citations
19 citations
Cites background or methods from "Arc Consistency Projection: A New G..."
...The AC-projection operator [14] differs from these redefinitions of graph and subgraph isomorphism [7,8] regarding the following aspects....
[...]
...The approach suggested in [14] advocates a projection operator based on the arc consistency algorithm....
[...]
...Indeed, these approaches use properties of the AC-projection operator initially introduced in [14]....
[...]
8 citations
Cites methods from "Arc Consistency Projection: A New G..."
...Recently, an approach related to ours has been introduced [15] in which arcconsistency was used for structuring the space of graphs....
[...]
4 citations
Cites background or methods from "Arc Consistency Projection: A New G..."
...Then, we present the AC-projection operator initially introduced in [3], as well as its very interesting properties....
[...]
...The approach suggested in [3] advocates a projection operator based on the arc consistency algorithm....
[...]
...In [3], the author introduced an interesting projection operator named AC-projection which seems to have good properties and ensure polynomial time and memory consumption....
[...]
...In this paper, we have studied the use of a new polynomial projection operator named AC-Projection initially introduced in [3] and based on a key technique of constraint programming namely Arc Consistency (AC)....
[...]
3 citations
Additional excerpts
...Terminons avec Liquière (2007) qui a proposé une nouvelle relation de généralité pour les graphes, cette relation n’étant pas toujours équivalente à la θ-subsomption....
[...]
...Cependant, nous avons établi que nos algorithmes dédiés bénéficiaient d’une meilleure complexité que les algorithmes généraux pour les graphes proposés par Liquière (2007)....
[...]
References
4,757 citations
"Arc Consistency Projection: A New G..." refers methods in this paper
...We use a new generality relation named AC-projection based on the arc consistency algorithm [12] and we prove that the search space, for a relational machine learning classification problem, is a concept lattice [13]....
[...]
...With this knowledge, we can define the notion of concept [13]....
[...]
3,204 citations
Additional excerpts
...For conceptual graphs [1] this operation is named projection....
[...]
1,527 citations
916 citations
560 citations
"Arc Consistency Projection: A New G..." refers background in this paper
...In machine learning, the complexity of this operation has motivated the use of learning biases: syntactic biases (trees [8], specific graph [9,10]), efficient implementation [7] and approximation of θ-subsumption [11]....
[...]
...But, thanks to AC-projection, we don’t have to explore all the elements of this set as in classical tree mining method [8]....
[...]