Post Reply 
(HP15C)(HP67)(HP41C) Bernoulli Polynomials
08-29-2023, 05:47 PM (This post was last modified: 08-29-2023 08:24 PM by Albert Chan.)
Post: #5
RE: (HP15C)(HP67)(HP41C) Bernoulli Polynomials
(08-29-2023 11:47 AM)Namir Wrote:  When I saw the nested summation formula in Wikipedia, I decided that it would be an easier approach to code
(albeit at the expense of some extra CPU time).

That's quite true.
But noted that formula had huge cancellation issue with factor (x+k)^m, when x and/or m is big.

[Image: 9c975ba4c30fbfcb6511fa4a623be8dfb7cb3422]

For example, if m=6, n=m (for last sum), it may have huge cancellation errors.

\(\displaystyle \frac{\Delta^6(x^6)}{7} =
\frac{x^6-6\,(x\!+\!1)^6+15\,(x\!+\!2)^6-20\,(x\!+\!3)^6+15\,(x\!+\!4)^6-6\,(x\!+\!5)^6+(x\!+\!6)^6}{7}
= \frac{720}{7} \)

Example, tested on XCas 1.9.0, with 48 bits float:

XCas> B(m,x) := sum(1/(n+1) * sum((-1)^k * comb(n,k) * (x+k)^m, k = 0 .. n), n = 0 .. m)

XCas> float(B(16, pi))      → 1462871.932         /* reference */
XCas> B(16, float(pi))      → -423965.265625    /* massive cancellations */
XCas> B2(16, float(pi))    → 1462871.93177      /* falling factorial form more stable */
Find all posts by this user
Quote this message in a reply
Post Reply 


Messages In This Thread
RE: (HP15C)(HP67)(HP41C) Bernoulli Polynomials - Albert Chan - 08-29-2023 05:47 PM



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