[
https://issues.apache.org/jira/browse/MATH-699?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sébastien Brisard updated MATH-699:
-----------------------------------
Attachment: inv-cum-new-impl.zip
Hi,
please find attached a proposal for a new implementation of
{{inverseCumulativeDistribution}}. For the time being, this new implementation
is implemented as a static (utility) class, as I would like you to check the
logics before committing it.
Highlights are
* use of Chebyshev's inequality (thanks Phil) to bracket the root
* in cases where the random variable has no mean or no variance, and is not
compactly supported, the method falls back to geometric bracketing, instead of
arithmetic bracketing.
* removal of calls to {{getDomainLowerBound(p)}}, {{getDomainUpperBound(p)}},
{{getInitialDomain(p)}}.
The attached class has a main() method which runs some performance checks (new
impl vs old impl). The report on my machine (Intel Core i5 M540 @2.53GHz, 4Go
RAM) tends to indicate that the new implementation is slightly quicker than the
old one.
Commented out is also a plateau detection which proves fairly robust (although
I'm sure we can fail it), but induces a slight decrease of the overall
performances. I'm just wondering if we should set a flag in the constructor, to
indicate whether or not plateaux should be detected ? Alternatively, we could
remove plateau detection altogether, and clearly state in the javadoc that the
default impl *fails* with plateaux.
Looking forward to reading your comments,
Sébastien
> inverseCumulativeDistribution fails with cumulative distribution having a
> plateau
> ---------------------------------------------------------------------------------
>
> Key: MATH-699
> URL: https://issues.apache.org/jira/browse/MATH-699
> Project: Commons Math
> Issue Type: Bug
> Affects Versions: 3.0
> Reporter: Sébastien Brisard
> Assignee: Sébastien Brisard
> Priority: Minor
> Attachments: AbstractContinuousDistributionTest.java,
> inv-cum-new-impl.zip
>
>
> This bug report follows MATH-692. The attached unit test fails. As required
> by the definition in MATH-692, the lower-bound of the interval on which the
> cdf is constant should be returned. This is not so at the moment.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
