[
https://issues.apache.org/jira/browse/NUMBERS-38?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16036254#comment-16036254
]
Amey Jadiye edited comment on NUMBERS-38 at 6/5/17 7:02 PM:
------------------------------------------------------------
Hi [~erans]
My initial thoughts was the formula given in code itself is the marker to test
and is perfectly written (as present in apache's project), your comment forced
me to verify everything from roots which ended up me reading lots of papers and
watching youtube on Gamma functions and approximations _(including great
interview of [Cornelius Lanczos|https://www.youtube.com/watch?v=avSHHi9QCjA])_
and finally I found our Code is Buggy generating wrong values for inputs :(
_(design bug not code!)_
There are the some references I cross verified the test input/outputs _(both
produce same result and our code produce something else)_
- [https://rosettacode.org/wiki/Gamma_function#Java]
- [https://en.wikipedia.org/wiki/Lanczos_approximation] (_this is python but
generates same outputs_)
I saw coefficient are take from (which are standard anyway), no matter which
set you take results are approx same.
- [https://mrob.com/pub/ries/lanczos-gamma.html]
For more cross verification there is online gamma calculator which also gives
same result.
- [http://www.efunda.com/math/gamma/findgamma.cfm]
All that I said, our code need to be fixed and it should generate
{{gamma(z)=(n-1)!}} (n minus one factorial) values.
let me know your thoughts.
was (Author: ameyjadiye):
Hi [~erans]
My initial thoughts was the formula given in code itself is the marker to test
and is perfectly written (as present in apache's project), your comment forced
me to verify everything from roots which ended up me reading lots of papers and
watching youtube on Gamma functions and approximations _(including great
interview of [Cornelius Lanczos|https://www.youtube.com/watch?v=avSHHi9QCjA])_
and finally I found our Code is Buggy generating wrong values for inputs :(
There are the some references I cross verified the test input/outputs _(both
produce same result and our code produce something else)_
- [https://rosettacode.org/wiki/Gamma_function#Java]
- [https://en.wikipedia.org/wiki/Lanczos_approximation] (_this is python but
generates same outputs_)
I saw coefficient are take from (which are standard anyway), no matter which
set you take results are approx same.
- [https://mrob.com/pub/ries/lanczos-gamma.html]
For more cross verification there is online gamma calculator which also gives
same result.
- [http://www.efunda.com/math/gamma/findgamma.cfm]
All that I said, our code need to be fixed and it should generate
{{gamma(z)=(n-1)!}} (n minus one factorial) values.
let me know your thoughts.
> No unit tests for "LanczosApproximation" class
> ----------------------------------------------
>
> Key: NUMBERS-38
> URL: https://issues.apache.org/jira/browse/NUMBERS-38
> Project: Commons Numbers
> Issue Type: Test
> Reporter: Gilles
> Labels: unit-test
> Fix For: 1.0
>
> Attachments: NUMBERS-38.patch
>
>
> The computation of the {{LanczosApproximation}} (package
> {{o.a.c.numbers.gamma}} in module {{commons-numbers-gamma}}) function is not
> checked by unit tests.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
