Learning Multiplicity Automata from Smallest Counterexamples
EuroCOLT '99 Proceedings of the 4th European Conference on Computational Learning Theory
Using Multiplicity Automata to Identify Transducer Relations from Membership and Equivalence Queries
ICGI '08 Proceedings of the 9th international colloquium on Grammatical Inference: Algorithms and Applications
A spectral learning algorithm for finite state transducers
ECML PKDD'11 Proceedings of the 2011 European conference on Machine learning and knowledge discovery in databases - Volume Part I
A finite model construction for coalgebraic modal logic
FOSSACS'06 Proceedings of the 9th European joint conference on Foundations of Software Science and Computation Structures
Myhill-Nerode theorem for sequential transducers over unique GCD-Monoids
CIAA'04 Proceedings of the 9th international conference on Implementation and Application of Automata
A spectral algorithm for learning Hidden Markov Models
Journal of Computer and System Sciences
| Hi-index | 0.00 |
It is shown that a real-valued function f(x), defined for strings x over a finite alphabet,is of the form (@bg(x)+@c) exp(@d|x|) for constants @b, @c, @d, and the acceptance probability function g for a probabilistic automation, if and only if f is of finite rank, where the latter external criterion is equivalent to the internal realizability of f by a finite-state sequential system permitted to have arbitrary real initial, transition, and output weights. The development encompasses multiple numerical outputs (finite vectors of functions of strings) and the corresponding generalization of this theorem; as an intermediate step, a set of sufficient conditions is established for equivalence of sequential systems (ss) with multiple outputs, yielding procedures for conversion of ss to numerical-output probabilistic automata (npa). Additional instances are given of application of these ideas in constructing npa equivalent to certain ss.