Encoding Nondeterministic Finite-State Tree Automata in Sigmoid Recursive Neural Networks
Recently, a number of authors have explored the use of recursive recursive neural nets (RNN) for the adaptive processing of trees or tree-like structures. One of the most important language-theoretical formalizations of the processing of tree-structured data is that of finite-state tree automata (FSTA). In many cases, the number of states of a nondeterministic FSTA (NFSTA) recognizing a tree language may be smaller than that of the corresponding deterministic FSTA (DFSTA) (for example, the language of binary trees in which the label of the leftmost k-th order grandchild of the root node is the same as that on the leftmost leaf). This paper describes a scheme that directly encodes NFSTA in sigmoid RNN.
KeywordsInput Symbol Tree Automaton Tree Language Tree Transducer State Port
- 1.Rafael C. Carrasco and Mikel L. Forcada. Simple strategies to encode tree automata in analog recursive neural networks. IEEE Transactions on Knowledge and Data Engineering, 2000. Accepted for publication.Google Scholar
- 2.Rafael C. Carrasco, Mikel L. Forcada, M. Ángeles Valdés-Muñoz, and Ramón P. Neco. Stable encoding of finite-state machines in discrete-time recurrent neural nets with sigmoid units. Neural Computation, 12, 2000. In press.Google Scholar
- 6.R.C. Gonzalez and M.G. Thomason. Syntactical pattern recognition. Addison-Wesley, Menlo Park, CA, 1978.Google Scholar
- 8.Stefan C. Kremer. A Theory of Grammatical Induction in the Connectionist Paradigm. PhD thesis, Department of Computer Science, University of Alberta, Edmonton, Alberta, 1996.Google Scholar
- 11.J.W. Thatcher. Tree automata: An informal survey. In A.V. Aho, editor, Currents in the theory of computing. Prentice-Hall, Englewood-Cliffs, NJ, 1973.Google Scholar