For a finite measurable partition, the relative entropy splits into the entropy of the cell masses plus the average entropy of the conditioned pieces; images of the pieces can only lose entropy; refining a partition can only increase the entropy of the cell masses; and suitable partitions bring it arbitrarily close to the full relative entropy.
In the settings of The Real Numbers: Standing Notation and Background and Measure Spaces and the Lebesgue Integral: Standing Notation, the measure space of the latter being instantiated at each use by the measure space named, let be a measurable space, let and be probability measures on it, and suppose that has finite relative entropy with respect to . Finite measurable partitions and refinement are those of that definition, conditioned probability measures are those of that definition, the partition entropy is that of that definition, and image measures are those of claim 1 of Image Measures, Measures with Densities, and Change of Variables. In claims 1 and 2, is a finite measurable partition of , and for , and the pieces of and of on are the conditioned measures whenever , and whenever .
1. (Chain rule) For every with one has , so that is defined; for every such the piece has finite relative entropy with respect to the piece ; and
2. (Images of the pieces) Let be a measurable space and measurable with respect to and . Then for every with , so that by claim 1, the measure has finite relative entropy with respect to , and
3. (Refinement) Let and be finite measurable partitions of such that refines ; the partition entropies over and over are defined by claim 1. Then .
4. (Approximation) For every positive there is a finite measurable partition of with .
Loading…
No relations recorded yet.