Questions tagged [kernels]
The kernels tag has no summary.
108 questions
3
votes
0
answers
150
views
Which RKHSs can be embedded in a local Sobolev space?
Let ${\cal F}_k$ be the RKHS of functions on an open set $\cal X\subseteq {\bf R}^n$ with kernel $k$. For which $k$ can ${\cal F}_k$ be embedded in the Sobolev space $W_\text{loc}^{\beta,2}(\cal X)$ (...
- 93
-1
votes
1
answer
175
views
Can we make assumptions such that $K(\cdot, x) \ll \lambda K$ for a.e. $x$ where $K$ is a markov kernel
Let $(X, \Sigma_X)$ and $(Y, \Sigma_Y)$ be two measurable spaces. Let $K: X \times \Sigma_Y \rightarrow [0,1]$ be a markov kernel. Let $\lambda$ be a $\sigma$-finite measure on $X$ and let $\lambda K (...
- 177
0
votes
0
answers
85
views
Rational embeddings of positive definite kernels over the natural numbers? (Unit vectors with rational coordinates)
I am searching for rational embeddings of positive definite kernels $k$ symmetric, taking rational values and $k(n,n)=1$ such as $k(a,b) = \min(a,b)/\max(a,b)$ or $k(a,b) = 2\gcd(a,b)/(a+b)$ or $k(a,...
0
votes
0
answers
64
views
Taking limits of Gaussian Processes - call for references
I'm investigating a particular topic and I'd like to get some references on it.
The idea is as follows: pick some natural $d$ and let $\mathcal{F}_d$ be a Gaussian Process on $\mathbb{R}^d$ with mean ...
- 103
1
vote
0
answers
95
views
MAGMA question (kernel of ring hom)
I apologize for posting here, but I don't see much else on stackoverflow where people who know MAGMA might be able to answer. The MAGMA documentation at
https://magma.maths.usyd.edu.au/magma/...
- 3,596
0
votes
0
answers
109
views
When is −∥x−y∥+∥A(x−y)∥ a CPD kernel ? - Difference of CPD radial kernels remains CPD (Bernstein functions)
I’m studying a family of translation-invariant kernels built from two norms and would like to understand exactly when their difference remains conditionally positive-definite. In particular, we focus ...
- 1
1
vote
1
answer
122
views
Positive definiteness of negative exponential of a matrix-valued quadratic form
For a matrix $M$ denote by $M^*$ its Hermitian conjugate.
For integers $d,m \geq 1$ consider the function $f:\mathbb{R}^m \to \mathbb{C}^{d \times d}$ defined for a column vector $\boldsymbol{x}=(x_1,\...
- 309
0
votes
0
answers
69
views
Inner Product in the RKHS of a Matrix-Valued Kernel
I am working with a matrix-valued reproducing kernel
$$ K: \mathbb{R}^d \times \mathbb{R}^d \to \mathbb{R}^{d \times d}, $$
which defines a corresponding reproducing kernel Hilbert space (RKHS) $H_K^d$...
- 101
1
vote
0
answers
86
views
How computationally efficient are kernel tricks? [closed]
"If we compare to non-kernel polynomial regression it is O(Tnp) where is p is dimension of polynomial while kernel polynomial is O(n^2d) + O(T*n^2) where d is original number of attributes, ...
3
votes
1
answer
277
views
Literature request: Covariance operators for Gaussian measures
I am looking to answer the question:
If $\mathcal{B}$ is a separable Banach space and $R: \mathcal{B}^*\to\mathcal{B}$ is a symmetric and positive operator, then $\phi: \mathcal{B}^*\to\mathbb{R}, \...
- 181
6
votes
1
answer
378
views
Non-equivalent definitions of Markov process
As far as I know, there are three definitions of Markov processes (or of Markov chains).
DEFINITION 1 (WEAKER). A process $(X_t)_{t\in[0,\infty)}$ on $(\Omega,\mathcal{F},\mathbb{P})$ with values in ...
- 1,590
2
votes
1
answer
159
views
Heat kernel convergence when expanding domains
Let $\Omega$ be an arbitrary domain in $\mathbb{R}^n$. There exists a positive $C^{\infty}$ function $G_{\Omega} : \Omega \times \Omega \times (0, \infty) \rightarrow \mathbb{R}$ (Dirichlet heat ...
- 687
4
votes
2
answers
270
views
Results of invertibility of a matrix involving the Szego kernel
In the context of reproducing kernel Hilbert spaces, the Szego kernel is the function $k(z_i,z_j)=\frac{1}{1-z_j\overline{z_j}}$.
Given $2n$ points $\{z_1,\ldots,z_n\},\{w_1,\ldots,w_n\}\in\mathbb{D}\...
- 277
1
vote
1
answer
209
views
Characterization of Fellerian kernels
This question concerns Feller Markov kernels, similar to Vanessa's question.
Terminology
By 'Markov kernel' $N:E\to F$, we adopt exactly the same definition as Vanessa, with the exception that $E,F$ ...
- 235
2
votes
0
answers
74
views
Characterization of Gaussian Gram matrices
From Euclidean geometry we know that a matrix $C$ is a matrix of squared Euclidean distances between some points if and only if $-\frac{1}{2} H D H \succeq 0$ (positive semi-definite) with $H = (I - \...
- 607
2
votes
1
answer
118
views
Non-linear transforms of RKHS question
I was reading the paper Norm Inequalities in Nonlinear Transforms (referenced in this question) but ran into difficulties, so I was wondering if anyone could help?
I think I follow the paper until I ...
- 41
2
votes
1
answer
105
views
Conditions for absorption
Let $X$ be a Markov chain with countable state space $S$ and transition kernel $P$, and let $h \colon S \to [0,1]$ be a sub-harmonic or super-harmonic function. Assume that for all $\varepsilon >0$ ...
user514580
1
vote
0
answers
193
views
Estimator for the conditional expectation operator with convergence rate in operator norm
Let $X$ and $Z$ be two random variables defined on the same probability space, taking values in euclidian spaces $E_X$ and $E_Z$, with distributions $\pi$ and $\nu$, respectively.
Let $L^2(\pi)$ ...
- 151
6
votes
0
answers
290
views
Gaussian lower heat kernel bounds on non-convex bounded domain
I am looking for a proof the following theorem.
Let $U \subset \mathbb{R}^n$ be a bounded domain with $C^2$ boundary and $p(x,y,t)$ be the Neumann heat kernel. Then there exist a constant $C>0$ ...
- 61
2
votes
1
answer
214
views
Orthonormal bases in RKHSs via interpolating sequences
Definitions and setting
Let $\mathcal{H}$ be a separable, infinite-dimensional, reproducing kernel Hilbert space on a nonempty set $X$. As usual, denote the reproducing kernel on $\mathcal{H}$ by $K$ ...
- 4,842
1
vote
0
answers
188
views
Gradient estimate of Dirichlet Heat kernel (Classical Laplacian)
Let $p^D(t,x,y)$ be the heat kernel for the Dirichlet Laplacian in an open set $D$. Do we have the following estimate and where can I find it ? $$\lvert\nabla_xp^D(t,x,y)\rvert\le C\dfrac{1}{\min (\...
17
votes
3
answers
1k
views
A kernel 'more analytic' than $\exp(-x^2)$
I am looking for an analytic function $F: \mathbb{R} \rightarrow (0,\infty)$ with $\int_{\mathbb{R}} F(x) \, dx = 1$ and the property, that $\sum\limits_{k=0}^{\infty} |c_k| \varepsilon^k (2k)! < \...
- 1,389
2
votes
1
answer
121
views
Pair of positive harmonic functions with negative inner product in Drury-Arveson space
Define a reproducing kernel on the Euclidean ball in $\mathbb{C}^d$ by
$$k(z,w)=\frac{1}{1-\langle z,w\rangle}+\frac{1}{1-\langle w,z\rangle}-1.$$
Call the corresponding real reproducing kernel ...
- 1,503
5
votes
1
answer
319
views
Is $k(A, B) = \text{Tr}[(A^{1/2} B A^{1/2})^{1/2}]$ a positive definite kernel?
Let $\mathbb{S}_n$ denote the set of $n \times n$ symmetric positive semidefinite matrices. I am trying to figure out whether $k: \mathbb{S}_n \times \mathbb{S}_n \to \mathbb{R}_+$ defined as:
$$k(A, ...
- 244
2
votes
0
answers
95
views
RKHS lying in another RKHS
Suppose $H_1$ and $H_2$ are reproducing kernel Hilbert spaces such that $H_1 \subset H_2$. For $f \in H_1$, when can I bound $\|f \|_1$ with $C\|f\|_2$ (for some $C$)? Is there a relationship between ...
- 93
1
vote
2
answers
108
views
Monotonicity of kernel matrices with respect to hyperparameters
Let $\mathcal{X}$ be some nice space, let $\Phi$ be some ordered space, and let $K :\mathcal{X} \times \mathcal{X} \times \Phi \to \mathbf{R}$ be a positive-semidefinite kernel indexed by a ...
- 949
2
votes
0
answers
576
views
Conditions for equivalence of RKHS norm and $L^2(P)$ norm
Let $K$ be a psd kernel on an abstract space $X$ and let $H_K$ be the induced Reproducing Kernel Hilbert Space (RKHS). Let $P$ be a probability measure on $X$ such that $H_K \subseteq L^2(P_X)$ and ...
- 7,043
1
vote
0
answers
107
views
What is lost after RKHS embedding of the L1 space?
We know that every distribution or $L^1$ function $f$ over space $\mathcal{X}$ (e.g., $R^d$) can be embedded to an RKHS $\mathcal{H}$ with a $1$-bounded kernel $\mathcal{K}$ (e.g., the RBF kernel) ...
- 43
1
vote
0
answers
130
views
$L_1$ convergence rates for multivariate kernel density estimation
Let $X$ be a random variable on $\mathbb R^d$ with probability density function $f$, and let $X_1,\ldots,X_n$ of $X$ be $n$ iid copies of $X$. Given a bandwidth parameter $h=h_n > 0$ and a kernel $...
- 7,043
4
votes
2
answers
326
views
Sharp Dirichlet heat kernel estimates in exterior domains?
I am right now working on some linear parabolic problems studying the behaviour of its solutions for large initial data. To do this, I have needed to use some estimates of the Dirichlet and Neumann ...
- 91
1
vote
0
answers
100
views
Properties of a kernel convolution $K'(x,y) = \int_X\int_X K_0(x,a)K(a,b)K_0(b,y)d\mu(a)d\mu(b)$ where $K$ and $K_0$ are kernels on $(X,\mu)$
Let $(X,\mu)$ be a probability measure space and $K:X \times X \to \mathbb R$ be a (psd) kernel on $X$. Let $K_0$ be another kernel on $X$ and defined a new kernel $\widetilde K$ on $X$ by
$$
\...
- 7,043
2
votes
1
answer
197
views
Representer theorem for a loss / functional of the form $L(h) := \sum_{i=1}^n (|h(x_i)-y_i|+t\|h\|)^2$
Let $K:X \times X \to \mathbb R$ be a (positive-definite) kernel and let $H$ be the induced reproducing kernel Hilbert space (RKHS). Fix $(x_1,y_1),\ldots,(x_n,y_n) \in X \times \mathbb R$. For $t \ge ...
- 7,043
4
votes
0
answers
1k
views
Reproducing kernel Hilbert space of Matérn kernels
I am trying to read a recent paper titled "Interpolation and learning with scale dependent kernels" by Pagliana, Ruidi, De Vito, and Rosasco. (The paper can be found on ArXiv)
On the top of ...
0
votes
0
answers
233
views
Is the kernel $\vert d_X - d_Y \vert^p$ conditionally negative definite?
Given two finite metric spaces $(X,d_X)$ and $(Y,d_Y)$, for $p > 0$, define the kernel ($4$-D tensor) $K$ on $(X \times Y)^2$ by:
$$K\big( (x_i, y_k), (x_j, y_l) \big) = \vert d_X(x_i, x_j) - d_Y(...
- 111
0
votes
1
answer
870
views
Proof: If a reproducing kernel exists for a Hilbert space, then it is unique
I really want to prove the statement in the title but I'm struggling with it. Here my current state:
Proof via contradiction. Let $\mathcal{H}$ be a RKHS with two reproducing kernels $k$ and $\hat{k}$ ...
- 13
4
votes
1
answer
545
views
Minimum upper bound for sum of the entries of the inverse covariance matrix
Let $x \in \mathbb{R}^n$ and $k$ is RBF kernel
$$k(x, x') := \exp \left(-\frac{\|x-x'\|^2}{2\sigma^2}\right)$$
and let $\mathbf{K}$ be the following $n \times n$ covariance matrix
$$\mathbf{K} = \...
3
votes
0
answers
469
views
Extending Ky Fan's eigenvalues inequality to kernel operators
--Migrating from MSE since it might fit better here--
Base result
The following result in Terry Tao's book, 'p. 47, Ky Fan inequality' reads as:
$$\sum_i\lambda_i(A+B) \leq \sum_i \lambda_i(A) + \...
- 277
0
votes
0
answers
274
views
A new method for processing music scores?
I have developed a method and python script:
https://github.com/githubuser1983/algorithmic_python_music
which allows the user to input a midi file and then chose a few numbers as parameters, and the ...
7
votes
1
answer
628
views
Why is this nonlinear transformation of an RKHS also an RKHS?
I came across this paper (beginning of page 6) where they stated that if $f,h\in \mathcal{H}$, where $\mathcal{H}$ is an RKHS, then $l_{h,f}=\left|f(x)-h(x)\right|^q$ where $q\geq 1$ also belongs to ...
- 393
1
vote
0
answers
175
views
Subspace of RKHS generated by kernel mean embeddings
Suppose $\mathcal{H_k}$ is a reproducing kernel Hilbert space (RKHS) with reproducing kernel $k: \mathcal{X} \times \mathcal{X} \rightarrow \mathbb{R}$. I am looking for results characterising the ...
- 93
4
votes
0
answers
157
views
What is the native Hilbert space associated with the kernel $\frac{\sum \min{(x_i,y_i)}}{\sum \max{(x_i,y_i)}}$?
In this answer on MSE it is shown that the function
$$ K:(\mathbb{R}^{>0})^n\times (\mathbb{R}^{>0})^n\rightarrow\mathbb{R}\,\quad K(x,y)=\frac{\sum_{i=1}^n\min{(x_i,y_i)}}{\sum_{i=1}^n\max{(x_i,...
- 316
2
votes
0
answers
98
views
Johnson filtration and lower central series
Let $G$ be a group and consider the lower central series:
$$G=\gamma_1 G \geq \gamma_2 G=[G,\gamma_1 G]\geq \gamma_3G=[G,\gamma_2G]\geq\cdots.$$
Let $S_g^1$ be a compact oriented genus $g$ surface ...
- 439
1
vote
1
answer
446
views
Simple example of Hammerstein integral equation
I'm currently reading this paper (and working on a similar one). The main goal is to study the Hammerstein integral equation (in $\mathcal{C}(I,E))$:
$$x(t) = \int_{0}^{t} K(t,s)f\big(s,x(s)\big)ds,\...
- 301
8
votes
0
answers
381
views
Monotonicity of log determinant of Gaussian kernel matrix
Let \begin{equation} k({x},{y}) = \sigma \exp\left(-\frac{(x-y)^2}{2\theta^2}\right)\end{equation}
be a squared-exponential (Gaussian) kernel, with $\sigma,\vartheta>0$. Consider, for a set of $N$ ...
- 55
3
votes
0
answers
571
views
Analytic formula for the eigenvalues of kernel integral operator induced by Laplace kernel $K(x,x') = e^{-c\|x-x'\|}$ on unit-sphere in $\mathbb R^d$
Let $d \ge 2$ be an integer and let $X=\mathcal S_{d-1}$ the unit-sphere in $\mathbb R^d$. Let $\tau_d$ be the uniform distribution on $X$. Define a function $K:X \times X \to \mathbb R$ by $K(x,y) := ...
- 7,043
1
vote
0
answers
270
views
Variance-based localized Rademacher complexity for RKHS unit-ball
Let $\mathscr X$ be a compact subset of $\mathbb R^d$ (e.g the unit-sphere). Let $K: \mathscr X \times \mathscr X \to \mathbb R$ be a positive kernel function and let $\mathscr H_K$ be the induced ...
- 7,043
1
vote
0
answers
153
views
Concentration of random Rayleigh quotients
Let $K$ be some large square matrix of height $N$, and $u$ a column vector of height $N$. Fixing $n$, take a random set of $n$ indices (with replacement) uniformly from $1,...,N$. Let $y$ be the ...
- 693
3
votes
1
answer
457
views
Is the solution to the Fredholm integral equation of the first kind a continuous analogue of Cramer's rule for matrix equations?
When we have a system of of $n$ linear equations represented by $$A \vec{x} = \vec{b} $$ with $\vec{x} = (x_{1}, x_{2}, \dots, x_{n})^{\intercal} $, we can solve for each component of this vector by ...
- 5,065
2
votes
0
answers
129
views
Kernels with finite dimensional feature spaces
Suppose $x,y \in \mathbb{R}^n$ for some given fixed n.
Consider a kernel $K(x,y) = f(\langle x, y \rangle)$, I'd like to know which functions $f$ admit a finite dimensional feature map. In other words,...
- 93
1
vote
0
answers
149
views
When do the kernels of module homomorphisms between rings whose kernels contain a given fixed ideal contain every prime ideal over it?
$\DeclareMathOperator{\Hom}{Hom}$All our rings are commutative with unity and, if necessary, we can suppose that they are actually polynomial rings over a field in finitely many variables where the ...
- 583