Finite relative entropy gives noise-connectedness to the Gaussian reference and a density, so the entropy domain lies in the set of the tangency theorem and inherits its noise map property.
Each result cited is universally quantified over the data in its own statement.
Claim 1. Let . The data of Noise-Optimal Maps out of the Entropy Domain of a Diagonal Gaussian Reference Measure are present: the reference measure is , is positive with for every , and has finite relative entropy with respect to . Hence by Noise-Optimal Maps out of the Entropy Domain of a Diagonal Gaussian Reference Measure §connected.
Claim 2. Let be the set of the that have a density with respect to , as defined in the statement of Tangency of Noise-Optimal Displacements out of a Measure with a Density Relative to a Diagonal Gaussian Measure, and the Noise Map Property, whose data are present here with this and the reference measure . Let . By Relative Entropy of Probability Measures §relative-entropy, has a density with respect to , and by claim 1; so . Thus .
By Tangency of Noise-Optimal Displacements out of a Measure with a Density Relative to a Diagonal Gaussian Measure, and the Noise Map Property §map-property, has the noise map property: for every and every the ordered pair is uniquely noise-mapped and for a noise-optimal map from to , where is the identity map of and the noise tangent space at , as The Noise Map Property of a Set of Probability Measures §map-property requires. That condition is imposed on each member of the set separately, so it holds in particular for every ; and by claim 1. Hence has the noise map property.
Loading…