Symbolic and connectionist learning techniques for grammatical inference

摘要: This thesis is structured in four parts for a total of ten chapters. The first part, introduction and review (Chapters 1 to 4), presents an extensive state-of-the-art both symbolic connectionist GI methods, that serves also state most the basic material needed describe later contributions thesis. These constitute contents rest 5 10). second on techniques regular grammatical inference 7), describes related theory methods GI, which include other lateral subjects such as representation oi. finite-state machines (FSMs) recurrent neural networks (RNNs). third part thesis, augmented expressions their inductive inference, comprises Chapters 8 9. (or AREs) are defined proposed new subclass CSLs does not contain all context-free languages but large class capable describing patterns with symmetries (context-sensitive) structures interest pattern recognition problems. fourth just includes Chapter 10: conclusions future research. 10 summarizes main results obtained points out lines further research should be followed deepen some theoretical aspects raised facilitate application developed tools real-world problems area computer vision.