Higher Order Derivatives in Costa's Entropy Power Inequality
Fan Cheng, Yanlin Geng

TL;DR
This paper investigates higher order derivatives of the differential entropy in Costa's EPI, revealing new convexity properties of Fisher information and proposing conjectures on the monotonicity of derivatives.
Contribution
It establishes the nonnegativity of the third and fourth derivatives of entropy, introduces conjectures on higher derivatives, and provides geometric insights into covariance-preserving transformations.
Findings
Third derivative of entropy is nonnegative, implying Fisher information is convex.
Fourth derivative of entropy is nonpositive, indicating specific curvature properties.
Proposes conjectures on the sign of higher order derivatives and the convexity of log Fisher information.
Abstract
Let be an arbitrary continuous random variable and be an independent Gaussian random variable with zero mean and unit variance. For , Costa proved that is concave in , where the proof hinged on the first and second order derivatives of . Specifically, these two derivatives are signed, i.e., and . In this paper, we show that the third order derivative of is nonnegative, which implies that the Fisher information is convex in . We further show that the fourth order derivative of is nonpositive. Following the first four derivatives, we make two conjectures on : the first is that is nonnegative in if is odd,…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsStatistical Mechanics and Entropy · Probabilistic and Robust Engineering Design · Mathematical Inequalities and Applications
