P. Dupont. Incremental regular inference. In Proceedings of the 3rd International Colloquium on Grammatical Inference (ICGI-96): Learning Syntax from Sentences, volume 1147 of LNAI, pages 222--237, Berlin, September 1996. Springer.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....on a correct grammar. The implemented algorithm just uses two specialisation operations, substitution of the non terminal on the left hand side of a production rule and deletion of a production rule, but clearly more operations could be added within this framework. Another hybrid algorithm is Dupont s (1996), which uses a combination of splitting and merging to adjust a regular grammar to each positive or negative example as it arrives; the grammar has to be kept consistent with all the patterns seen previously, so the algorithm has to store the sets of positive and negative examples seen so far (the ....

Dupont, P., 1996, Incremental regular inference. Lecture Notes in Artificial Intelligence, 1147: 222--237.

....avoiding the wumpus. Because the strategies of the wumpus can be simply represented as a finite state machine. The method of learning the strategies of the wumpus was to learn a DFA representation of the strategies. Researchers have shown that learning the DFA can be done both incrementally (Dupont 1994) and PAC learnably (Parekh and Hanovar 1994) Because the DFA 1 Copyright 1998, American Association for Artificial Intelligence (www.aaai.org) All rights reserved. can be learned quickly and incrementally, it makes for an excellent choice in learning the wumpus strategies in an unsupervised, ....

Dupont, P. 1994. Incremental Regular Inference. In Miclet, L., and Higuera, C., eds., Proceedings of the Third ICGI-96, Lecture Notes in Artificial Intelligence 1147, 222-237, Montpellier, France: Springer.

....Xg: S is the initial state and the set of nal states is F = fSg. 3.2. Learning nite automata One of the most famous algorithm inferring nite automata from positive and negative instances rpni, has been proposed in [22, 23] Although this work has been since rened or supplemented by many authors [14, 1, 10, 15, 16], the basic method of applying a state merging technique in Lat(PTA(I ) is widely used. We propose thus to implement the behavior of rpni in a logic programming environment. Basically, rpni starts from a current automaton A = PTA(I ) and while possible, merges pairs of states in A. It consider ....

P. Dupont. Incremental regular inference. In L. Miclet and C. de la Higuera, editors, Grammatical Inference: Learning Syntax from Sentences, ICGI'96, volume 1147 of Lecture Notes in Articial Intelligence, pages 222237. Springer Verlag, September 1996.

....of new canonical objects, namely the automata A(L; f ) This also simpli es correctness proofs of inference algorithms for k reversible languages, k 0, to some extent. It seems to be interesting to study these canonical automata also in the search space framework of Dupont and Miclet [4, 6, 5]. We feel that deterministic methods (such as the one proposed in this paper) are quite important for practical applications, since they could be understood more precisely than mere heuristics, so that one can prove certain properties about the algorithms. Moreover, the approach of this paper ....

P. Dupont. Incremental regular inference. In L. Miclet and C. de la Hieguera, editors, Proceedings of the Third International Colloquium on Grammatical Inference (ICGI-96): Learning Syntax from Sentences, volume 1147 of LNCS/LNAI, pages 222-237, Springer, 1996.

....has been processed. Though provably correct, this algorithm has practical limitations because the size of the lattice grows exponentially with the number of states of the PTA. Dupont has proposed an incremental version of the RPNI algorithm [Oncina Garc ia, 1992] for regular grammar inference [Dupont, 1996]. This algorithm is also based on the idea of a lattice of partitions of the states of a PTA for a set of positive examples. It uses information from a set of negative examples to guide the ordered search through the lattice and is guaranteed to converge to the target DFA when the set of examples ....

....on the teacher to provide accurate responses to membership queries poses a potential limitation in applications where a reliable teacher is not available. We are exploring the possibility of learning in an environment where the learner does not have access to a teacher. The algorithms due to Dupont [Dupont, 1996] and Porat Feldman [Porat Feldman, 1991] operate in this framework. Some open problems include whether the limitations of these algorithms (e.g. need for storage of all the previously seen examples and complete lexicographic ordering of examples) can be overcome without sacrificing efficiency ....

Dupont, P. (1996). Incremental Regular Inference. Pages 222--237 of: Miclet, L., & Higuera, C. (eds), Proceedings of the Third ICGI-96, Lecture Notes in Artificial Intelligence 1147. Montpellier, France: Springer.

....been processed. Though provably correct, this algorithm has practical limitations because the size of the lattice grows exponentially with the number of states of the PTA. The incremental version of the regular positive and negative inference (RPNI) algorithm [OG92] for regular grammar inference [Dup96] is also based on the idea of a lattice of partitions of the states of a PTA for a set of positive examples. It uses information from a set of negative examples to guide the ordered search through the lattice. It requires storage of all the examples seen by the learner to ensure that each time the ....

....the teacher to provide accurate responses to membership queries poses a potential limitation in applications where a reliable teacher is not available. We are exploring the possibility of learning in an environment where the learner does not have access to a teacher. The algorithms due to Dupont [Dup96] and Porat Feldman [PF91] operate in this framework. Some open problems include whether the limitations of these algorithms (e.g. need for storage of all the previously seen examples and complete lexicographic ordering of examples) can be overcome without sacrificing efficiency and guaranteed ....

P. Dupont. Incremental regular inference. In L. Miclet and C. Higuera, editors, Proceedings of the Third ICGI-96, Montpellier, France, Lecture Notes in Artificial Intelligence 1147, Springer-Verlag, pages 222--237, 1996.

....identification of a DFA consistent with a given set S = S [ S Gamma . If the sample is a characteristic set for the target DFA then the algorithm is guaranteed to return a canonical representation of the target DFA. Our description of the RPNI algorithm is based on the explanation given in [8]. A labeled sample S = S [ S Gamma is provided as input to the algorithm. It constructs a prefix tree automaton PTA(S ) and numbers its states in the standard order. Then it performs an ordered search in the space of partitions of the set of states of PTA(S ) under the control of ....

....For example, if q i , q j , and q k are states of M such that for some a 2 Sigma, ffi(q i ; a) fq j ; q k g then the states q j and q k are merged together. This recursive merging of states can go on for at most N Gamma 1 steps and the resulting automaton M is guaranteed to be a DFA [8]. Note that since we know by the grammar covers relation that if M accepts a negative example in S Gamma then so would M . The function, consistent(M ; S Gamma ) returns True if M is consistent with all examples in S Gamma and False otherwise. If a partition is ....

[Article contains additional citation context not shown here]

P. Dupont. Incremental regular inference. In L. Miclet and C. Higuera, editors, Proceedings of the Third ICGI-96, Lecture Notes in Artificial Intelligence 1147, pages 222--237, Montpellier, France, 1996.

No context found.

P. Dupont. Incremental regular inference. In Proceedings of the 3rd International Colloquium on Grammatical Inference (ICGI-96): Learning Syntax from Sentences, volume 1147 of LNAI, pages 222--237, Berlin, September 1996. Springer.

No context found.

P. Dupont. Incremental regular inference. In L. Miclet and C. Higuera, editors, Proceedings of the Third ICGI-96, Lecture Notes in Artificial Intelligence 1147, pages 222--237, Montpellier, France, 1996.

No context found.

Dupont, P., Incremental Regular Inference, in: Grammatical Inference: Learning Syntax from Sentences, LNAI 1147, 1996, pp. 222--237.

No context found.

P. Dupont. Incremental regular inference. In L. Miclet and C. Higuera, editors, Proceedings of the Third ICGI-96, Lecture Notes in Artificial Intelligence 1147, pages 222--237, Montpellier, France, 1996. Springer.

No context found.

DUPONT, P. Incremental regular inference. In: (ICGI'96) Lecture Notes in Artificial Intelligence, n. 1147, Springer Verlag, pages 222-237, 1996.

No context found.

P. Dupont. Incremental regular inference. In Grammatical Inference: Learning Syntax from Sentences, ICGI'96, number 1147 in Lecture Notes in Artificial Intelligence, pages 222-237. Springer Verlag, 1996.

No context found.

Dupont, P. (1996a). Incremental Regular Inference. Pages 222--237 of: Miclet, L., & Higuera, C. (eds), Proceedings of the Third ICGI-96, Lecture Notes in Artificial Intelligence 1147. Montpellier, France: Springer.

No context found.

Intelligence 28(2):127--162. Dupont, P. 1996. Incremental regular inference. Lecture Notes in Computer Science 1147:222.

No context found.

Dupont, P. (1996a). Incremental Regular Inference. Pages 222--237 of: Miclet, L., & Higuera, C. (eds), Proceedings of the Third ICGI-96, Lecture Notes in Artificial Intelligence 1147. Montpellier, France: Springer.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC