Attribute Grammars for Genetic Representations of Neural Networks and Syntactic Constraints of Genetic Programming

An attribute grammar (Knuth, 1968; Bochmann, 1976) is a context-free grammar augmented by the assignment of semantic attributes to the symbols of the grammar. A production rule specifies not only the replacement of symbols, but also the evaluation of the symbol’s attributes. In our research, an attribute grammar is used to specify classes of neural network structures with explicit representation of their functional organization. These representations provide useful constraints upon a genetic optimization that guarantee the preservation of syntactically correct genetic trees with semantically meaningful sub-trees. In this paper, we give a broad overview of our research into attribute grammar representations, from the basic and known capabilities, to the current ideas being addressed, to the future directions of our research.