Post Reply 
HP Prime Numerical Integration
08-03-2022, 06:15 PM
Post: #1
HP Prime Numerical Integration
Hello,
I'd like to know the method of numerical integration the hp prime uses, because it's great.
For example, I tried to compute the incomplete elliptic integral of the first kind:

EllipticF(x,m) = integrate 1/sqrt(1 - msin^2(t)) dt from 0 to x
For parameters x=2, m=3, so:
EllipticF(2,3) = integrate 1/sqrt(1 - 3sin^2(t)) dt from 0 to 2

Wolfram Alpha tells me that the result is:
EllipticF(2,3) = 1.001077-1.490278*i

On HP Prime the command would be:

int(1/SQRT(1-3*(SIN(T)^2)),T,0,2)

When approximated in HP Prime this results in 1.00107718634 - 1.49027809758*i

When I tried the same thing in Wolfram Alpha:

N[integrate (1/sqrt(1-3sin^2(t)) dt from 0 to 2)]
the result is 0.9996911316703074 - 1.4848975811276912*i

Then I tried other methods, such as programming my own methods in python. I tried the Romberg integration, Gauss Legandre Quadrature, Simpson Rule, Gauss Kronrod.
None of which worked as good. Even romberg in HP Prime can't calculate it well enough.

romberg(1/sqrt(1-3*(sin(t)^2)),t,0,2) returns the last approximation 0.991198869818-1.47660568998*i

The scipy.integrate.quad in python works well enough, but it's not as good as HP Prime int.
Since 1/sqrt(1-3sin^2(x)) has poles from 0 to 2, I suspect that some pole finding algorithm might be used.
Find all posts by this user
Quote this message in a reply
Post Reply 


Messages In This Thread
HP Prime Numerical Integration - Ruda975 - 08-03-2022 06:15 PM



User(s) browsing this thread: 1 Guest(s)