Posts: 1,396

Threads: 340

Joined: Feb 2009

Mick posted an intresting question here :

http://math.stackexchange.com/questions/...x-fx-x1t-1
It reminds of the binary partition function we discussed before which was strongly related by the similar equation F ' (x) = F(x/2).

Maybe the method to solve F ' (x) = F(x/2) can be used/modified to solve Mick's differential equation.

Anyway I think its intresting.

regards

tommy1729

Posts: 1,396

Threads: 340

Joined: Feb 2009

Although far from an answer , using the Mittag-Leffler function to Get a fake (1+x)^t will probably get us a good approximation in terms of a Taylor series.

Although getting these Taylor coëfficiënts is Nice , its not a closed form asymtotic.

Unless we get a simple rule for these coef , its hard to get THE asym from the Taylor.

Regards

Tommy1729

Posts: 1,396

Threads: 340

Joined: Feb 2009

04/10/2015, 03:57 PM
(This post was last modified: 04/10/2015, 04:08 PM by tommy1729.)
Clearly this function grows slower than any exponential but faster than any polynomial or even exp(x^t).

This implies that if our asymptotic is entire - or the function itself ?? - then it is determined by its zero's completely.

Otherwise fake function theory can be applied.

And then the fake is compl determined by its zero's.

A quick brute estimate is exp(x^T + T x^{tT}) , where T is between t and 1.

ITS not accurate i know.

All of this is ofcourse Up to a multip constant.

Regards

Tommy1729

Posts: 1,396

Threads: 340

Joined: Feb 2009

About post 2 i use truncated carleman matrices for (fake) entire functions.

And then simply solve THE equation.

Regards

Tommy1729