Date(s) : 04/09/2014 iCal
14 h 00 min - 15 h 00 min
ABSTRACT: Weighted automata provide a concise algebraic parametrization for functions from strings to real numbers. This class contains many well-known\nexamples like deterministic finite automata (DFA) — where values are\nbinary — and hidden Markov models (HMM) — where values represent\nprobabilities of strings. In this talk I will present a general framework\nbased on weighted automata which can be used to tackle a wide variety of\nlearning problems involving sequential data\, including classification\,\ndensity estimation\, and sequence tagging. I will then show how recent\nspectral algorithms for learning stochastic languages and sequence tagging\nmodels can be derived naturally within this framework.
Catégories Pas de Catégories