arXiv subjects

Otte Heinävaara

Publications and source records attributed to Otte Heinävaara.

7 recordsLinked to original sources

arXiv2026

Convolution comparison measures

We give a precise functional comparison between classical and free convolutions. If $μ$ and $ν$ are compactly supported probability measures, we show that the expectation of $f$ over the classical convolution $μ* ν$ is at least the expectation of $f$ over the free convolution $μ\boxplus ν$, as long as the fourth derivative of $f$ is non-negative. Conversely, the non-negativity of the fourth derivative is necessary for such a comparison. This comparison is based on the positivity of a related measure on $\mathbb{R}^{2}$, which we dub the convolution comparison measure. We give an expression for this measure using a curious identity involving Hermitian matrices.

math.FA↗

arXiv2023

Tracial joint spectral measures

Given two Hermitian matrices, $A$ and $B$, we introduce a new type of spectral measure, a $\textit{tracial joint spectral measure}$ $μ_{A, B}$ on the plane. Existence of this measure implies the following two results: 1) any two-dimensional subspace of the Schatten-$p$ class is isometric to a subspace of $L_{p}$, and 2) if $f : \mathbb{R} \to \mathbb{R}$ has non-negative $k$th derivative and $A$ and $B$ are Hermitian matrices with $A$ positive semidefinite, then $t \mapsto \mathrm{tr}~f(t A + B)$ has non-negative $k$th derivative. We also give an explicit expression for the measure $μ_{A, B}$.

math.FA↗

arXiv2023

A random line intersects $\mathbb{S}^2$ in two probabilistically independent locations

We consider random lines in $\mathbb{R}^3$ (random with respect to the kinematic measure) and how they intersect $\mathbb{S}^2$. It is known that the entry point and the exit point behave like \textit{independent} uniformly distributed random variables. We give a new proof using bilinear integral geometry and use this approach to show that this property is extremely rare: if $K \subset \mathbb{R}^n$ is a bounded, convex domain with smooth boundary with this property (i.e., the intersection points with a random line are independent), then $n=3$ and $K$ is a ball.

math.PR↗

arXiv2022

Planes in Schatten-$3$

We prove that any two-dimensional real subspace of Schatten-3 can be linearly isometrically embedded into $L_{3}$. This resolves the case $p = 3$ of Hanner's inequality for Schatten classes, conjectured by Ball, Carlen and Lieb. We conjecture that similar isometric embedding is possible for any $p \geq 1$.

math.FA↗

arXiv2019

Characterizing matrix monotonicity of fixed order on general sets

We give new characterizations for matrix monotonicity and convexity of fixed order which connects previous characterizations by Loewner, Dobsch, Donoghue, Kraus and Bendat--Sherman. The ideas introduced are then used to characterize matrix monotone functions of arbitrary order on general subsets of the real line.

math.FA↗

arXiv2017

Local characterizations for the matrix monotonicity and convexity of fixed order

We establish local characterizations of matrix monotonicity and convexity of fixed order by giving integral representations connecting the Loewner and Kraus matrices, previously known to characterize these properties, to respective Hankel matrices. Our results are new already in the general case of matrix convexity and our approach significantly simplifies the corresponding work on matrix monotonicity. We also obtain an extension of the original characterization for matrix convexity by Kraus, and tighten the relationship between monotonicity and convexity.

math.FA↗

arXiv2016

On the inconsistency of $\ell_1$-penalised sparse precision matrix estimation

Various $\ell_1$-penalised estimation methods such as graphical lasso and CLIME are widely used for sparse precision matrix estimation. Many of these methods have been shown to be consistent under various quantitative assumptions about the underlying true covariance matrix. Intuitively, these conditions are related to situations where the penalty term will dominate the optimisation. In this paper, we explore the consistency of $\ell_1$-based methods for a class of sparse latent variable -like models, which are strongly motivated by several types of applications. We show that all $\ell_1$-based methods fail dramatically for models with nearly linear dependencies between the variables. We also study the consistency on models derived from real gene expression data and note that the assumptions needed for consistency never hold even for modest sized gene networks and $\ell_1$-based methods also become unreliable in practice for larger networks.

cs.LG↗