Physical Review Letters · 1996 · 51 citations · 13 references
ComputingComputer ScienceMachine LearningStochastic ModelingStochastic ProcessesBayesian InferenceMarkov Chain Monte CarloOnline AlgorithmsPattern RecognitionProbabilistic Machine LearningReinforcement LearningBayesian Online LearningGibbs DistributionMini-batch LearningMini-batch TrainingOffline Reinforcement LearningOnline Bayesian InferenceOnline Machine LearningPosterior Distribution Approximation
I propose a general model of on-line learning from random examples which, when applied to a smooth realizable stochastic rule, yields the same asymptotic generalization error rate as optimal batch algorithms. The approach is based on an iterative Gaussian approximation to the posterior Gibbs distribution of rule parameters.
13
The statistical mechanics of learning a rule
Timothy L. H. Watkin, Albrecht Rau, Michael Biehl · Reviews of Modern Physics · 1993 · 473 citations · Full text