Y. Mukouchi, Inductive inference of an approximate concept from positive data, in: Proc. 4th Internat. Conf. on Algorithmic Learning Theory, LNAI 872 (SpringerVerlag, 1994) 484-499.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....languages [1, 36, 37, 43] Mukouchi and Arikawa [28] showed that the class of length bounded EFSs is also refutable. This notion is a new criterion introduced by them that a learner can refute each hypothesis space if it turns out to be insucient for identi cation. Many other researchers such as [20, 21, 26, 27] enjoyed various topological properties of EFSs on inductive inference. Jain and Sharma [18] analyzed the mind change complexity and the intrinsic complexity of EFSs. In contrast to the learnability of EFSs on inductive inference, the polynomial time learnability is another interesting theme on ....

Y. Mukouchi, Inductive inference of an approximate concept from positive data, in: Proc. 4th Internat. Conf. on Algorithmic Learning Theory, LNAI 872 (SpringerVerlag, 1994) 484-499.

....on bounded unions of pattern languages [1, 31, 32, 41] Mukouchi and Arikawa [25] showed that for a hierarchy of hypothesis spaces LB EFS(m) for m 0, there exists a learner who can refute each hypothesis space if it turns out to be insufficient for identification. Moriyama, Sato [23] Mukouch [24] and Kobayashi, Yokomori [17] showed that classes of EFS enjoy various good topological properties in inductive inference. Kobayashi [18] also used EFS as a tool for uniformly showing the efficient inferability of a collection of language classes in identification in the limit. Sugimoto et al. ....

Y. Mukouchi, Inductive inference of an approximate concept from positive data, in: Proc. the 4th Int'l. Conf. on Algorithmic Learning Theory, LNAI 872 (Springer-Verlag, 1994) 484--499.

.... of the current hypothesis space, and change it to a newer one that may contain the target [5] Mukouchi also presented a framework of approximate identification of a concept in the limit, where it is only required for the learning device to output a minimal concept containing the target concept [6]. This paper concerns the latter approach. For studying on the approximate learning, it is important to exactly define the closeness of an output of the learning device to a target concept. In this paper, we consider a problem of identifying a best fit approximation of a target concept with ....

Y. Mukouchi. Inductive Inference of an Approximate Concept from Positive Data, in Proc. of 5th International Workshop on Algorithmic Learning Theory, Lecture Notes in Artificial Intelligence 872, Springer-Verlag, pp.484-499, 1994

....in L j is uniformly decidable for all indices j. As a matter of fact, the definition of an indexed family contains both, a description for every enumerated language L j and a particular enumeration of all the languages from its range. Recently, this topic has attracted much attention (cf. e.g. [24, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36, 37, 43, 45, 50, 51]) Next we specify the information from which the target languages have to be learned. A text of a language L is an infinite sequence of strings that eventually contains all strings of L. Since a text contains exclusively positive examples concerning the target language, we sometimes refer to ....

....family itself as hypothesis space. If an indexed family L has to be inferred with respect to the hypothesis space L, then we refer to this learning model as to proper inference. Note that proper inference has been studied by various authors (cf. e.g. Angluin [1, 2] Shinohara [45] Mukouchi [36]) Nevertheless, one may also allow any recursive enumeration of the range of L as well as any description of the enumerated languages provided membership remains uniformly decidable. The resulting learning model is referred to as class preserving inference. Moreover, when dealing with learning ....

Y. Mukouchi, Inductive inference of an approximate concept from positive data, Kyushu University, Research Institute of Fundamental Information Science, Technical Report RIFIS-TR-CS 74, 1993.

....a learner with the ability to refute the hypothesis space, in case that a target concept is not contained in it. Another one is permitting a learner to output an approximate concept, such as a minimal concept, when a target concept is outside the hypothesis space, which was recently studied by [Muk94] and [Sak91] using the framework of Gold s identification in the limit from positive data [Gol67] This paper closely concerns the latter approach. On the other hand, some interesting works attempting to generalize the Valiant s Probably Approximately Correct (or PAC) learning model [Val84] have ....

....is a C upper approximation of Tg is finite. An indexed family C has M finite thickness iff C satisfies both MEF and MFF conditions. Mukouchi presented an interesting result on a sufficient condition for the upper approximate identifiability from positive data using the following lemma. Lemma 8. [Muk94] Let C be an indexed family which satisfies MEF condition and has finite elasticity, let L U be a nonempty concept, and let Ln 2 C be a concept. a) If L Ln , then there exists a C upper approximation L j 2 C of L such that L j Ln . b) If Ln is a C upper approximation of L, then there ....

[Article contains additional citation context not shown here]

Y. Mukouchi. Inductive Inference of an Approximate Concept from Positive Data, in Proc. of 5th International Workshop on Algorithmic Learning Theory, Lecture Notes in Artificial Intelligence 872, Springer-Verlag, pp.484-499, 1994

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