Ryo Yoshinaka

Radu Cosmin Dumitru, Ryo Yoshinaka, Ayumi Shinohara

It is well known that computing a minimum deterministic finite automaton consistent with a given set of positive and negative examples is NP-hard. Previous work has identified conditions on the input sample under which the problem becomes tractable or remains hard. In this paper, we study the computational complexity of the case where the input sample is prefix-closed. This formulation is equivalent to computing a minimum Moore machine consistent with observations along its runs. We show that the problem is NP-hard to approximate when the sample set consists of all prefixes of binary strings. Furthermore, we show that the problem remains NP-hard as a decision problem even when the sample set consists of the prefixes of a single binary string. Our argument also extends to the corresponding problem for Mealy machines.

Kaizaburo Chubachi, Diptarama Hendrian, Ryo Yoshinaka et al.

We propose a query learning algorithm for residual symbolic finite automata (RSFAs). Symbolic finite automata (SFAs) are finite automata whose transitions are labeled by predicates over a Boolean algebra, in which a big collection of characters leading the same transition may be represented by a single predicate. Residual finite automata (RFAs) are a special type of non-deterministic finite automata which can be exponentially smaller than the minimum deterministic finite automata and have a favorable property for learning algorithms. RSFAs have both properties of SFAs and RFAs and can have more succinct representation of transitions and fewer states than RFAs and deterministic SFAs accepting the same language. The implementation of our algorithm efficiently learns RSFAs over a huge alphabet and outperforms an existing learning algorithm for deterministic SFAs. The result also shows that the benefit of non-determinism in efficiency is even larger in learning SFAs than non-symbolic automata.

Takeaki Uno, Hiroki Maegawa, Takanobu Nakahara et al.

We address the problem of un-supervised soft-clustering called micro-clustering. The aim of the problem is to enumerate all groups composed of records strongly related to each other, while standard clustering methods separate records at sparse parts. The problem formulation of micro-clustering is non-trivial. Clique mining in a similarity graph is a typical approach, but it results in a huge number of cliques that are of many similar cliques. We propose a new concept data polishing. The cause of huge solutions can be considered that the groups are not clear in the data, that is, the boundaries of the groups are not clear, because of noise, uncertainty, lie, lack, etc. Data polishing clarifies the groups by perturbating the data. Specifically, dense subgraphs that would correspond to clusters are replaced by cliques. The clusters are clarified as maximal cliques, thus the number of maximal cliques will be drastically reduced. We also propose an efficient algorithm applicable even for large scale data. Computational experiments showed the efficiency of our algorithm, i.e., the number of solutions is small, (e.g., 1,000), the members of each group are deeply related, and the computation time is short.