Acquisition of Multiple Tree Structured Patterns by an Evolutionary Method Using Sets of Tag Tree Patterns as Individuals

Using Label Information in a Genetic Programming Based Method for Acquiring Tag Tree Patterns with Vertex Labels and Wildcards

“…represents a wildcard for a vertex or an edge label. A tag tree pattern with vertex and edge labels and wildcards is a tree structured pattern that has edge and vertex labels and is allowed to have wildcards for edge labels and wildcards for vertex labels, which is an extension of a tag tree pattern with only edge labels and wildcards used in the previous work [7]. A tag tree pattern with vertex and edge labels and wildcards is simply called a tag tree pattern with wildcards, or a tag tree pattern.…”

“…Genetic Programming (GP) is an evolutionary method that is an extension of Genetic Algorithm (GA) and deals with tree structured objects such as formulas and programs [1,4]. Obtaining characteristic tree structured patterns such as tag tree patterns, using Genetic Programming, from glycan data are proposed [2,3,9,7]. Tag tree patterns are tree structured patterns with structured variables that can be substituted by tree structured data.…”

“…In our previous work [7], we use a tag tree pattern with only edge labels and wildcards for edge labels and converted glycan data with only edge labels. In this paper, we extend the representation of a tag tree pattern in our previous work.…”

“…In order to obtain highly characteristic tag tree patterns, by using label information of positive examples, that is, label connecting relation of positive tree examples, we propose a learning method for acquiring characteristic tag tree pattern with wildcards using Genetic Programming, from positive and negative tree structured data. The proposed method is considered an extension of GP approach [2,7] for acquiring characteristic tag tree patterns from positive and negative tree structured data. As the glycan data g 1 and g 2 in Figure 2 show, the structure of a glycan is abstractly represented as a tree by representing single sugars as vertices and covalent bonds between sugars as edges.…”

“…The learning method of the previous work [2,7] generates tag tree patterns without checking label information of positive examples in GP operations. The method can generate tag tree patterns that match little or no positive examples in a given example set.…”