共查询到20条相似文献,搜索用时 31 毫秒
1.
We continue the investigation of some problems in learning theory in the setting formulated by F. Cucker and S. Smale. The
goal is to find an estimator
on the base of given data
that approximates well the regression function
of an unknown Borel probability measure
defined on
We assume that
belongs to a function class
It is known from previous works that the behavior of the entropy numbers
of
in the uniform norm
plays an important role in the above problem. The standard way of measuring the error between a target function
and an estimator
is to use the
norm (
is the marginal probability measure on X generated by
). This method has been used in previous papers. We continue to use this method in this paper. The use of the
norm in measuring the error has motivated us to study the case when we make an assumption on the entropy numbers
of
in the
norm. This is the main new ingredient of thispaper. We construct good estimators in different settings: (1) we know both
and
; (2) we know
but we do not know
and (3) we only know that
is from a known collection of classes but we do not know
An estimator from the third setting is called a universal estimator. 相似文献
2.
Denote by
the real-linear span of
, where
Under the concept of left-monogeneity defined through the generalized
Cauchy-Riemann operator we obtain the direct sum decomposition of
where
is the right-Clifford module of finite linear combinations of functions of the form
, where, for
, the function R is a k- or
-homogeneous leftmonogenic
function, for
or
, respectively, and h is a function defined in [0,∞) satisfying a certain integrability condition in relation to k, the spaces
are invariant under Fourier transformation.
This extends the classical result for
. We also deduce explicit Fourier transform
formulas for functions of the form
refining Bochner’s formula for spherical k-harmonics. 相似文献
3.
Ilia Krasikov 《Constructive Approximation》2008,28(2):113-125
T. Erdelyi, A.P. Magnus and P. Nevai conjectured that for
the orthonormal Jacobi polynomials
satisfy the inequality
[Erdelyi et al., Generalized Jacobi weights, Christoffel functions, and Jacobi polynomials. SIAM J. Math. Anal., 25 (1994),
602-614.]. Here we will confirm this conjecture in the ultraspherical case
even in a stronger form by giving very explicit upper bounds. We also show that
for a certain choice of
such that the interval
contains all the zeros of
Slightly weaker bounds are given for polynomials of odd degree. 相似文献
4.
Zoltan Furedi 《Discrete and Computational Geometry》2007,38(2):273-288
Let
be a triangle and let
be a set of homothetic copies of
. We prove that
implies that there are positive and negative signs
and there exist translates of
that cover
. 相似文献
5.
An affine pseudo-plane X is a smooth affine surface defined over
which is endowed with an
-fibration such that every fiber is irreducible and only one fiber is a multiple fiber. If there is a hyperbolic
-action on X and X is an
-surface, we shall show that the universal covering
is isomorphic to an affine hypersurface
in the affine 3-space
and X is the quotient of
by the cyclic group
via the action
where
and
It is also shown that a
-homology plane X with
and a nontrivial
-action is an affine pseudo-plane. The automorphism group
is determined in the last section. 相似文献
6.
In this article we show that the distributional point values of a tempered distribution are characterized by their Fourier
transforms in the following way: If
and
, and
is locally integrable, then
distributionally if and only if there exists k such that
, for each a > 0, and similarly in the case when
is a general distribution. Here
means in the Cesaro sense. This result generalizes the characterization of Fourier series of distributions with a distributional
point value given in [5] by
. We also show that under some extra conditions, as if the sequence
belongs to the space
for some
and the tails satisfy the estimate
,\ as
, the asymmetric partial sums\ converge to
. We give convergence results in other cases and we also consider the convergence of the asymmetric partial integrals. We
apply these results to lacunary Fourier series of distributions. 相似文献
7.
Rostom Getsadze 《Journal of Fourier Analysis and Applications》2006,12(5):597-604
We prove the following theorem: For arbitrary
there exists a nonnegative
function
such that
and
almost everywhere on
where
is the double Walsh-Paley system.
This statement remains true also for the double trigonometric system. 相似文献
8.
We give sharp lower estimates for the partial sums of the Fourier series
with both an even and odd number of terms. Our results are obtained through a monotonicity property of their local minima.
In particular, we prove that all partial sums are positive on
. 相似文献
9.
10.
We show that every function in the Hardy space can be approximated by linear combinations of translates and dilates of a synthesizer
, provided only that
and
satisfies a mild regularity condition. Explicitly, we prove scale averaged approximation for each
,
where
is an arbitrary lacunary sequence (such as
) and the coefficients
are local averages of f. This formula holds in particular if the synthesizer
is in the Schwartz class, or if it has compact support and belongs to
for some
in terms of differences of
. 相似文献
11.
D.S. Lubinsky 《Constructive Approximation》2007,25(3):303-366
Assume
is not an integer. In papers published in 1913 and 1938,
S.~N.~Bernstein established the limit
Here
denotes the error in best uniform approximation of
by polynomials
of degree
. Bernstein proved that
is itself the error in best uniform approximation of
by entire functions of exponential type at most 1,
on the whole real line. We prove that the best approximating entire function
is unique, and satisfies an alternation property. We show that the scaled
polynomials of best approximation converge to this unique entire function.
We derive a representation for
, as well
as its
analogue for
. 相似文献
12.
Miodrag Zivkovic 《Semigroup Forum》2006,73(3):404-426
Let
be the set of all
Boolean matrices. Let R(A) denote the row space of
, let
, and let
. By extensive computation we found that
and therefore
. Furthermore,
for
. We proved that if
, then the set
contains at least
elements. 相似文献
13.
Jacek Dziubanski 《Constructive Approximation》2008,27(3):269-287
Let
be the standard Laguerre functions of type a. We denote
. Let
and
be the semigroups associated with the orthonormal systems
and
. We say that a function f belongs to the Hardy space
associated with one of the semigroups if the corresponding maximal function belongs to
. We prove special atomic decompositions of the elements of the Hardy spaces. 相似文献
14.
The aim of this paper is to study the well-posedness of the initial-boundary value problem
where
is a bounded regular open domain in
is the outward normal to
and
, where
are pairwise disjoint measurable subsets of
with respect to Lebesgue surface measure on
. The main novelty lies on the reactive dynamical boundary condition imposed on
. The technique makes it possible to study the more general initial-boundary value problem
where
is as before and
. A key step in our analysis consists in studying the eigenvalue problem
相似文献
15.
Let
denote the linear space over
spanned by
. Define the (real) inner product
, where V satisfies: (i) V is real analytic on
; (ii)
; and (iii)
. Orthogonalisation of the (ordered) base
with respect to
yields the even degree and odd degree orthonormal Laurent polynomials
, and
. Define the even degree and odd degree monic orthogonal Laurent polynomials:
and
. Asymptotics in the double-scaling limit
such that
of
(in the entire complex plane),
, and
(in the entire complex plane) are obtained by formulating the odd degree monic orthogonal Laurent polynomial problem as a
matrix Riemann-Hilbert problem on
, and then extracting the large-n behaviour by applying the non-linear steepest-descent method introduced in [1] and further
developed in [2],[3]. 相似文献
16.
In this paper we develop a robust uncertainty principle for
finite signals in
which states that, for nearly all choices
such that
there is no signal
supported on
whose discrete Fourier transform
is supported on
In fact, we can make the above uncertainty principle quantitative in the sense that if
is supported on
then only a small percentage of the energy (less than half, say) of
is concentrated on
As an application of this robust uncertainty principle (QRUP), we consider the problem of decomposing a signal into a sparse
superposition of spikes and complex sinusoids
We show that if a generic signal
has a decomposition
using spike and frequency locations in
and
respectively, and obeying
then
is the unique sparsest possible decomposition (all other decompositions have more nonzero terms). In addition, if
then the sparsest
can be found by solving a convex optimization problem. Underlying our results is a new probabilistic approach which insists
on finding the correct uncertainty relation, or the optimally sparse solution for nearly all subsets but not necessarily all
of them, and allows us to considerably sharpen previously known results [9], [10]. In fact, we show that the fraction of sets
for which the above properties do not hold can be upper bounded by quantities like
for large values of
The QRUP (and the application to finding sparse representations) can be extended to general pairs of orthogonal bases
For nearly all choices
obeying
where
there is no signal
such that
is supported on
and
is supported on
where
is the mutual coherence between
and
An erratum to this article is available at . 相似文献
17.
Let E be a compact subset of C. We prove that if E satisfies the following local Markov property: for each polynomial P,
where M, m, s are positive constants independent of P,
and
; then E is L-regular, i.e. regular in the sense of the potential theory. In particular, if
satisfies the global Markov inequality, then E is L-regular. We also prove that if
is an m-perfect set (there exists c > 0 such that, for all
and $r\in (0,1]$,
and
, then E is L-regular. Examples given by Siciak [20] show that the assumption that m < 2 cannot be omitted. 相似文献
18.
The digitisation
of a real disc
having radius
and centre
consists of all integer points inside
, i.e.,
In this paper we show that there are
different (up to translations) digitisations of discs having radius
. More formally,
The result is of interest in the area of digital image processing because it describes how large the impact of the object
position can be on its digitisation. 相似文献
19.
20.
We give conditions on radial nonnegative weights $W_1We give conditions on radial nonnegative weights
and
on
, for which the a priori inequality
holds with constant independent of
. Here
is the Laplace-Beltrami operator on the sphere
. Due to the relation between
and the tangential component of the gradient,
, we obtain some "Morawetz-type" estimates for
on
. As a consequence we establish some new estimates for the free Schr?dinger propagator
, which may be viewed as certain refinements of the
-(super)smoothness estimates of Kato and Yajima. These results, in turn, lead to the well-posedness of the initial value problem
for certain time dependent first order spherical perturbations of the
dimensional Schr?dinger equation. 相似文献