arXiv subjects

Jacob Carruth

Publications and source records attributed to Jacob Carruth.

11 recordsLinked to original sources

arXiv2024

The Sobolev extension problem on trees and in the plane

Let $V$ be a finite tree with radially decaying weights. We show that there exists a set $E \subset \mathbb{R}^2$ for which the following two problems are equivalent: (1) Given a (real-valued) function $\phi$ on the leaves of $V$, extend it to a function $\Phi$ on all of $V$ so that $||\Phi||_{L^{1,p}(V)}$ has optimal order of magnitude. Here, $L^{1,p}(V)$ is a weighted Sobolev space on $V$. (2) Given a function $f:E \rightarrow \mathbb{R}$, extend it to a function $F \in L^{2,p}(\mathbb{R}^2)$ so that $||F||_{L^{2,p}(\mathbb{R}^2)}$ has optimal order of magnitude.

math.FA↗

arXiv2024

Almost Optimal Agnostic Control of Unknown Linear Dynamics

We consider a simple control problem in which the underlying dynamics depend on a parameter $a$ that is unknown and must be learned. We study three variants of the control problem: Bayesian control, in which we have a prior belief about $a$; bounded agnostic control, in which we have no prior belief about $a$ but we assume that $a$ belongs to a bounded set; and fully agnostic control, in which $a$ is allowed to be an arbitrary real number about which we have no prior belief. In the Bayesian variant, a control strategy is optimal if it minimizes a certain expected cost. In the agnostic variants, a control strategy is optimal if it minimizes a quantity called the worst-case regret. For the Bayesian and bounded agnostic variants above, we produce optimal control strategies. For the fully agnostic variant, we produce almost optimal control strategies, i.e., for any $\varepsilon>0$ we produce a strategy that minimizes the worst-case regret to within a multiplicative factor of $(1+\varepsilon)$.

math.OC↗

arXiv2023

An example related to Whitney's extension problem for $L^{2,p}(\mathbb{R}^2)$ when $1<p<2$

In this paper, we prove the existence of a bounded linear extension operator $T: L^{2,p}(E) \rightarrow L^{2,p}(\mathbb{R}^2)$ when $1<p<2$, where $E \subset \mathbb{R}^2$ is a certain discrete set with fractal structure. Our proof makes use of a theorem of Fefferman-Klartag on the existence of linear extension operators for radially symmetric binary trees.

math.CA↗

arXiv2023

A Bounded Regret Strategy for Linear Dynamics with Unknown Control

We consider a simple linear control problem in which a single parameter $b$, describing the effect of the control variable, is unknown and must be learned. We work in the setting of agnostic control: we allow $b$ to be any real number and we do not assume that we have a prior belief about $b$. For any fixed time horizon, we produce a strategy whose expected cost is within a constant factor of the best possible.

math.OC↗

arXiv2023

Optimal Agnostic Control of Unknown Linear Dynamics in a Bounded Parameter Range

Here and in a follow-on paper, we consider a simple control problem in which the underlying dynamics depend on a parameter $a$ that is unknown and must be learned. In this paper, we assume that $a$ is bounded, i.e., that $|a| \le a_{\text{MAX}}$, and we study two variants of the control problem. In the first variant, Bayesian control, we are given a prior probability distribution for $a$ and we seek a strategy that minimizes the expected value of a given cost function. Assuming that we can solve a certain PDE (the Hamilton-Jacobi-Bellman equation), we produce optimal strategies for Bayesian control. In the second variant, agnostic control, we assume nothing about $a$ and we seek a strategy that minimizes a quantity called the regret. We produce a prior probability distribution $d\text{Prior}(a)$ supported on a finite subset of $[-a_{\text{MAX}},a_{\text{MAX}}]$ so that the agnostic control problem reduces to the Bayesian control problem for the prior $d\text{Prior}(a)$.

math.OC↗

arXiv2023

Controlling Unknown Linear Dynamics with Almost Optimal Regret

Here and in a companion paper, we consider a simple control problem in which the underlying dynamics depend on a parameter $a$ that is unknown and must be learned. In this paper, we assume that $a$ can be any real number and we do not assume that we have a prior belief about $a$. We seek a control strategy that minimizes a quantity called the regret. Given any $\varepsilon>0$, we produce a strategy that minimizes the regret to within a multiplicative factor of $(1+\varepsilon)$.

math.OC↗

arXiv2021

Controlling Unknown Linear Dynamics with Bounded Multiplicative Regret

We consider a simple control problem in which the underlying dynamics depend on a parameter that is unknown and must be learned. We exhibit a control strategy which is optimal to within a multiplicative constant. While most authors find strategies which are successful as the time horizon tends to infinity, our strategy achieves lowest expected cost up to a constant factor for a fixed time horizon.

math.OC↗

arXiv2021

The norm of linear extension operators for $C^{m-1,1}(\mathbb{R}^n)$

Fix integers $m\ge 2$, $n\ge 1$. We prove the existence of a bounded linear extension operator for $C^{m-1,1}(\R^n)$ with operator norm at most $\exp(\gamma D^k)$, where $D := \binom{m+n-1}{n}$ is the number of multiindices of length $n$ and order at most $m-1$, and $\gamma,k > 0$ are absolute constants (independent of $m,n,E$). Upper bounds on the norm of this operator are relevant to basic questions about fitting a smooth function to data. Our results improve on a previous construction of extension operators of norm at most $\exp(\gamma D^k 2^D)$. Along the way, we establish a finiteness theorem for $C^{m-1,1}(\R^n)$ with improved bounds on the involved constants.

math.FA↗

arXiv2018

A coordinate-free proof of the finiteness principle for the Whitney extension problem

We present a coordinate-free version of Fefferman's solution of Whitney's extension problem in the space $C^{m-1,1}(\mathbb{R}^n)$. While the original argument relies on an elaborate induction on collections of partial derivatives, our proof uses the language of ideals and translation-invariant subspaces in the ring of polynomials. We emphasize the role of compactness in the proof, first in the familiar sense of topological compactness, but also in the sense of finiteness theorems arising in logic and semialgebraic geometry.

math.CA↗

arXiv2017

The Beurling-Selberg Box Minorant Problem via Linear Programming Bounds

In this paper we investigate a high dimensional version of Selberg's minorant problem for the indicator function of an interval. In particular, we study the corresponding problem of minorizing the indicator function of the box $Q_{N}=[-1,1]^N$ by a function whose Fourier transform is supported in the same box $Q_N$. We show that when the dimension is sufficiently large there are no minorants with positive mass and we give an explicit lower bound for such dimension. On the other hand, we explicitly construct minorants for dimensions $1,2,3,4$ and $5$ and, as an application, we use them to produce an improved diophantine inequality for exponential sums.

math.CA↗

arXiv2012

A comparison of the discrete Kolmogorov-Smirnov statistic and the Euclidean distance

Goodness-of-fit tests gauge whether a given set of observations is consistent (up to expected random fluctuations) with arising as independent and identically distributed (i.i.d.) draws from a user-specified probability distribution known as the "model." The standard gauges involve the discrepancy between the model and the empirical distribution of the observed draws. Some measures of discrepancy are cumulative; others are not. The most popular cumulative measure is the Kolmogorov-Smirnov statistic; when all probability distributions under consideration are discrete, a natural noncumulative measure is the Euclidean distance between the model and the empirical distributions. In the present paper, both mathematical analysis and its illustration via various data sets indicate that the Kolmogorov-Smirnov statistic tends to be more powerful than the Euclidean distance when there is a natural ordering for the values that the draws can take -- that is, when the data is ordinal -- whereas the Euclidean distance is more reliable and more easily understood than the Kolmogorov-Smirnov statistic when there is no natural ordering (or partial order) -- that is, when the data is nominal.

stat.ME↗