A noise penalty pair is lambda-displacement convex if the penalty lies above its first-order expansion along every noise-optimal coupling out of a score-domain measure, with a lambda/2 times noise-cost correction.
In the setting of First-Order Equations on the Noise Wasserstein Space Relative to a Noise Penalty Pair: Standing Notation, let be a noise penalty pair on and let . For one has , so that is a real number, and the score lies in , hence in , both by Noise Penalty Pairs on the Noise Wasserstein Space: the Penalty, Its Score, and Their Domains §pair. For the ordered pair is noise-connected 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 Noise Penalty Pairs on the Noise Wasserstein Space: the Penalty, Its Score, and Their Domains §pair, so that noise-optimal couplings are defined, and for such the noise displacement pairing and the noise cost are the real numbers of First-Order Equations on the Noise Wasserstein Space Relative to a Noise Penalty Pair: Standing Notation §couplings.
(-displacement convexity) The noise penalty pair is -displacement convex if
for every , every and every noise-optimal coupling . It is displacement convex if it is -displacement convex.
Loading…
No relations recorded yet.