The Gaussian entropy pair has closed and bounded penalty sublevel sets in the noise Wasserstein space, and by Talagrand's inequality the squared distance to the reference measure is at most 2 kappa over beta times the penalty.
In the setting of A Diagonal Gaussian Reference Measure on the Noise Wasserstein Space, Rescaled Heads and Gaussian Tails: Standing Notation, so that the reference measure is , let and be positive real numbers with for every , and let be the Gaussian entropy pair with temperature , whose hypothesis holds with this ; it is a noise penalty pair on by The Gaussian Entropy Pair is a Noise Penalty Pair, with Nonnegative Penalty and Dense Score Domain §pair. Noise-closedness is read in the metric space of The Noise Wasserstein Distance is a Metric on the Measures Noise-Connected to the Reference Measure: Existence of Noise-Optimal Couplings, Comparison with the Quadratic Wasserstein Distance and Lower Semicontinuity §metric; the real number quantified as in Noise-Closed Noise Penalty Pairs §noise-closed is a bound variable there, distinct from the variance sequence . For the number is defined by The Noise Wasserstein Distance is a Metric on the Measures Noise-Connected to the Reference Measure: Existence of Noise-Optimal Couplings, Comparison with the Quadratic Wasserstein Distance and Lower Semicontinuity §connected, since by The Gaussian Entropy Pair on the Noise Wasserstein Space: Relative Entropy and the Noise Score Field §penalty-domain and by The Noise Wasserstein Distance is a Metric on the Measures Noise-Connected to the Reference Measure: Existence of Noise-Optimal Couplings, Comparison with the Quadratic Wasserstein Distance and Lower Semicontinuity §reference, and is the absolute value of .
1. (Noise-closed) The pair is noise-closed.
2. (Growth bound) For every ,
Loading…
No relations recorded yet.