R. Parekh and V. Honavar. Learning DFA from simple examples. Machine Learning, 44(1/2):9--35, 2001.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....for instance [19] are too easy (nearly anything can be learned) identification in the limit [20] Based on identification in the limit, we will use identification in the limit from polynomial time and data [6] as our learning model. Learning from queries [13] or simple PAC learning [21] are alternative frameworks. It should be noticed that if a class of grammars is identifiable in the limit from polynomial time and data, the simple grammars in this class are also simple PAC learnable [22] In this setting the learner is asked to learn from a learning sample, i.e. a finite set of ....

Parekh, R., Honavar, V.: Learning dfa from simple examples. Machine Learning Journal 44 (2001) 9--35

.... 2 K(xjc) Simple examples (i.e. of low conditional complexity K(xjc) are more probable under m c . A class of concepts is said to be learnable from simple examples, or PACS learnable in short, if it is PAC learnable provided the examples are drawn according to m c . It has been proved in (Parekh and Honavar, 1997) that the class REG is PACS learnable. Learning models with teaching sets raise a problem when a representation of the target concept can be encoded by means of examples. We show that in this case, every learning algorithm can be rendered trivial: indeed, nothing prevents the teacher to put an ....

....polynomial in 1= and l. Connections between GM learnability and PEC learnability have been studied in (Denis and Gilleron, 1997b) Castro and Guijarro have studied connections between this model and exact learning with queries (Castro and Guijarro, 1998) DFA have been proved PACS learnable in (Parekh and Honavar, 1997). Proposition 1. If a class of language L is polynomially GM learnable, then it is PEC learnable from simple examples. Proof. Sketch. Let T be a teaching algorithm. Since T runs in polynomial time, there exists a integer k such that for every language L 2 L, Card(T (L) size(L) k . Let r ....

Parekh, R. and Honavar, V. (1997). Learning DFA from simple examples. In Li, M. and Maruoka, A., editors, Proceedings of the 8th International Workshop on Algorithmic Learning Theory (ALT-97), volume 1316 of LNAI, pages 116-131, Berlin. Springer.

....study a distribution dependent form of PAC learning that uses probability distributions related to Kolmogorov complexity. We relate the PACS model, defined by Denis, D Halluin and Gilleron in [3] with the standard simple PAC model and give a general technique that subsumes the results in [3] and [6]. 1 Introduction One of the most relevant models of learning is PAC learning , introduced by Valiant [7] which has been widely used to investigate the phenomenon of learning from examples. Informally, in this model one has to learn a target concept with high probability in polynomial time (and, ....

....this framework, called PACS model, examples with short descriptions with respect to the target concept have high probability to be drawn. We show here that any query learnable class is also learnable under this extended framework. So, for example, learnability of DFA s in the PACS model (shown in [6]) is a consequence of its query learnability [1] We also study the relationships between simple PAC and PACS. We show that they are correlated if some conditions hold. These conditions are not unreasonable since they hold for all known algorithms. An interesting consequence of our results is that ....

[Article contains additional citation context not shown here]

R. Parekh and V. Honavar. "Learning dfa from simple examples". In Proc. of the 8 th International Workshop on Algorithmic Learning Theory, pages 116--131. Lecture Notes in Artificial Intelligence, 1361. Springer, 1997.

....is based on an Occam algorithm. We also deduce that k reversible languages are learnable with simple examples from positive data and DFA are learnable with simple examples when a bound on the number of states of the target DFA is known by the learner. Also, DFA are learnable with simple examples (Parekh and Honavar, 1997) when the length of the longest example is added to the set of parameters for measuring the time complexity. Our model is defined in Section 1. The Occam s razor theorem and its converse are given and proved in Section 2. Learnability of decision lists is proved in Section 3. Learnability of ....

....In this simple PAC learning model, the examples are drawn according to m c instead of m. The main difference is that not only classes of simple concepts are proved learnable in this model. For example, the class of DNF formulae is learnable in this model. DFA are proved learnable in this model (Parekh and Honavar, 1997) but the reader should note that, for this result, the length of examples is added to the set of parameters for measuring the time complexity of a learning algorithm. This is problematic because arbitrarily long examples may be drawn with the SolomonoffLevin distribution. Again, it is not clear ....

Parekh, B. and Honavar, V. (1997). Learning dfa from simple examples. volume 1316 of LNAI, pages 116--131, Berlin. Springer.

....(where m c (x) 2 GammaK(xjc) In a word, simple examples (i.e. of low conditional complexity K(xjc) are more probable under m c . A class of concepts is said to be learnable from simple examples if it is PAC learnable provided the examples are drawn according to m c . It has been proved in [PH97] that the class REG is learnable from simple examples. The learning models with teaching sets raises a problem when a representation of the target concept can be encoded by means of examples. We show that in this case, every learning algorithm can be rendered trivial: indeed, nothing prevents the ....

....A for L in R which runs in time polynomial in 1=ffi and l. Connections between GM learnability and PEC learnability have been studied in [DG97b] Castro and Guijarro have studied connections between this model and exact learning with queries [CG98] DFA have been proved PACS learnable in [PH97]. Proposition 1. If a class of language L is polynomially GM learnable, then it is PEC learnable from simple examples. Proof. Sketch. Let T be a teaching algorithm. Since T runs in polynomial time, there exists a integer k such that for every language L 2 L, Card(T (L) size(L) k . Let r be ....

B. Parekh and V. Honavar. Learning DFA from simple examples. volume 1316 of LNAI, pages 116--131, Berlin, 1997. Springer.

No context found.

R. G. Parekh and V. G. Honavar. Learning dfa from simple examples. In Proceedings of the Eighth International Workshop on Algorithmic Learning Theory (ALT'97), Lecture Notes in Artificial Intelligence 1316, pages 116--131, Sendai, Japan, 1997. Springer. Also presented at the Workshop on Grammar Inference, Automata Induction, and Language Acquisition (ICML'97), Nashville, TN, July 12, 1997.

No context found.

R. G. Parekh and V. G. Honavar. Learning dfa from simple examples. In Proceedings of the Eighth International Workshop on Algorithmic Learning Theory (ALT'97), Lecture Notes in Artificial Intelligence 1316, pages 116--131, Sendai, Japan, 1997.

No context found.

R. Parekh and V. Honavar. Learning DFA from simple examples. Machine Learning, 44(1/2):9--35, 2001.

No context found.

R. Parekh and V. Honavar, "Learning DFA from simple examples," in Workshop on Automata Induction, Grammatical Inference, and Language Acquisition, ICML-97, 1997.

No context found.

R. Parekh and V. Honavar. "Learning dfa from simple examples". In Proc. of the 8 International Workshop on Algorithmic Learning Theory, pages 116--131. Lecture Notes in Artificial Intelligence, 1361. Springer, 1997.

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