3.5.Β SequentialLearning.BayesStationaryEnv
Bayesian stationary environments
This file defines the structure IsBayesAlgEnvSeq and provides its basic properties.
Main definitions
-
IsBayesAlgEnvSeq Q ΞΊ alg E A Y P: states that there is a measureP : Measure Ξ©such that the parameterE : Ξ© β πhas lawQand that the sequences of actionsA : β β Ξ© β πand feedbacksY : β β Ξ© β π¨are generated by the algorithmalg : Algorithm π π¨interacting with an underlying environment that depends onEandΞΊ(stationaryEnv (ΞΊ.sectR (E Ο))). -
bayesTrajMeasure Q ΞΊ alg: for any choice of probability measureQ : Measure π, Markov kernelΞΊ : Kernel (π Γ π) π¨, and algorithmalg : Algorithm π π¨, provides a probability measureP : Measure (β β π Γ π Γ π¨)on a space that carriesE,A, andYsuch thatIsBayesAlgEnvSeq Q ΞΊ alg E A Y P. -
bayesTrajMeasurePosterior Q ΞΊ alg n: aKernel (Iic n β π Γ π¨) πthat represents the posterior overEgiven the history up to timenunder the priorQand the algorithmalg, assuming that the kernelΞΊspecifies howEgives rise to the underlying (stationary) environment. See alsoLeanMachineLearning/SequentialLearning/AlgorithmDensityBayes.lean.
Main results
-
ae_IsAlgEnvSeq h: ifh : IsBayesAlgEnvSeq Q ΞΊ alg E A Y P, forQ-almost everye : π,IsAlgEnvSeq A' Y' alg (stationaryEnv (ΞΊ.sectR e)) (condDistrib (trajectory A Y) E P e)for some sequence of actionsA' : β β (β β π Γ π¨) β πand sequence of feedbacksY' : β β (β β π Γ π¨) β π¨. Intuitively, if the observable trajectory is generated by an underlying parametere : π, the measure that carries theIsBayesAlgEnvSeqstructure reveals a measure that carries anIsAlgEnvSeqstructure under the environmentstationaryEnv (ΞΊ.sectR e)and the same algorithm. This allows transferring results from theIsAlgEnvSeqstructure to theIsBayesAlgEnvSeqstructure.
Module LeanMachineLearning.SequentialLearning.BayesStationaryEnv contains 17 exposed declarations.
-
Learning.IsBayesAlgEnvSeq -
Learning.IsBayesAlgEnvSeq.hasLaw_action_zero -
Learning.IsBayesAlgEnvSeq.hasCondDistrib_action' -
Learning.IsBayesAlgEnvSeq.hasCondDistrib_feedback' -
Learning.IsBayesAlgEnvSeq.hasLaw_IT_action_zero -
Learning.IsBayesAlgEnvSeq.hasCondDistrib_IT_feedback_zero -
Learning.IsBayesAlgEnvSeq.hasCondDistrib_IT_action -
Learning.IsBayesAlgEnvSeq.hasCondDistrib_IT_feedback -
Learning.IsBayesAlgEnvSeq.hasLaw_IT_hist -
Learning.IsBayesAlgEnvSeq.ae_IsAlgEnvSeq -
Learning.bayesStationaryEnv -
Learning.IsAlgEnvSeq.isBayesAlgEnvSeq -
Learning.IT.bayesTrajMeasure -
Learning.IT.instIsProbabilityMeasureForallNatProdBayesTrajMeasure -
Learning.IT.isBayesAlgEnvSeq_bayesTrajMeasure -
Learning.IT.bayesTrajMeasurePosterior -
Learning.IT.instIsMarkovKernelForallSubtypeNatMemFinsetIicProdBayesTrajMeasurePosterior