Statistical and Computational Tradeoffs in Stochastic Composite Likelihood

12th International Conference on Artificial Intelligence and Statistics

Joshua V. Dillon ; Guy Lebanon

PID  :  120859

来源: CEUR

【 摘 要 】

Maximum likelihood estimators are often of limited practical use due to the intensive computation they require. We propose a fam ily of alternative estimators that maximize a stochastic variation of the composite like lihood function. We prove the consistency of the estimators, provide formulas for their asymptotic variance and computational com plexity, and discuss experimental results in the context of Boltzmann machines and con ditional random fields. The theoretical and experimental studies demonstrate the effec tiveness of the estimators in achieving a pre defined balance between computational com