| Home / lib / M_Mathematics / | ||
Halasz G., Lovasz L., Simonovits M., Sos V.T. (eds.) Paul Erdos and his mathematics, vol.1 (Springer, 2002)(KA)(600dpi)(T)(734s)_M_.djvu |
|
Size 7.8Mb Date Jan 27, 2006 |
them in a somewhat organized and unified way...
If f : G-+B is such that \f{x + y) - f(x) - f{y)\ < К for every x,y E G
with a nonnegative constant К then there is a homomorphism H from G
into the additive group of В such that \f - H\ <K everywhere on G...
Clearly,
and
u/(/,z0) = max G(ж0) - /(a?o),/(so) - Д
In th§ sequel we shall investigate the difference property of the classes
Ck(G,B) = {/ : G -> В : u;(/,z0) < if for every z0 € G}
and
C#(G,B) = {/ : G -> В : о/(/,я0) < # for every x0 E G},
where G is an Abelian topological group and В is a Banach space...
Let p be a polynomial such that p(i) = f(i) for i = 0,1,2; p(l/2) >
/A/2); and pC/2) < /C/2)...
That is, if 0 < ai < /3i < 1 and 0 < a2 < /32 < 1 then ^
The problem, whether or not Па^ = (Fa) for every 0 < a < /3 < 1
remains open...
Suppose we can prove that g = g\ + H + 5, where
Pi is measurable, H is additive and the differences of S are null...
However, an application of 3.1 and 7.6 gives that the class of
those derivatives that satisfy | f(x) — /@)| < К for every x with a constant
K, has the difference property...
The proof of A) used capacity; in particular,
it was shown that with probability one a Brownian path has zero capacity
for d > 4 but positive capacity for d < 3...
In trying to
make a rigorous argument out of this heuristic, DEK discovered in fact that
with probability one the Brownian path in one dimension has no cut times...
As in the case of Brownian cut times in one dimension, there
was a heuristic argument using a random time which would indicate this
might give the right order of decay, but the random time was not a stopping
time...
It is not difficult to show that the Brownian exponent exists using
scaling and subadditivity...
More recently, it
was proved [17], with probability one, the set of cut times in two and three
dimensions is 1 - (&A, l)/2)...
Amongst his many deep contributions to mathe-
mathematics, this formula must have been the simplest...
That quadrature formulae should play a role in analysing Lagrange in-
interpolation is scarcely surprising, as the two topics are first cousins...
Continuity of / has been weakened
to Riemann integrability of / and that has been weakened to improper
Riemann integrability, thereby allowing / to have a number of infinities in
the interval...
They also strengthened the convergence A4) in terms of errors of best
approximation, defined above...
Indeed, in very sharp contrast, Erdos and
Vertesi [38], [39] showed in 1980 that there is always / € C[-l, 1] for which
} diverges a.e.:
Theorem of Erdos and Vertesi on a.e...
On the one hand, for specific weights, the range
of p admitting convergence has been established, and for general weights,
necessary conditions have been established...
In his work for Jacobi weights, Askey established the Marcinkiewicz
inequality A9) using the positivity of suitable order Cesaro means of the
{Sn}, as well as Jensen's inequality...
For Wfc)Q, it was shown in [54] that if we add to the zeros {^jn}j=1
of pn, the two points in (—1,1) at which |pn^Q| achieves its maximum,
both of which are close to ±1, then we can achieve a better result than that
above: the conclusion of (a) above holds for 1 < p < 00, so there is no need
to damp with 1 + Q^Q^a-
The reader will have doubtless observed that all the concrete positive
conclusions above deal with interpolation at zeros of orthogonal polynomi-
polynomials, or something very similar...
| © 2007 eKnigu | ||