[
https://issues.apache.org/jira/browse/NUMBERS-98?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16811878#comment-16811878
]
Gilles edited comment on NUMBERS-98 at 4/8/19 10:23 AM:
--------------------------------------------------------
Hi [~Ting223].
Thanks a lot for your interest in this project.
We certainly need to tackle the linear algebra issue; however the "how" is not
clear to me; then, somehow, this JIRA ticket escaped my attention, and we don't
have enough time left to discuss/research the implications and their impact on
your GSoC proposal...
Also because some aspects will require more opinions, the discussion must be
brought to the "dev" mailing list of the "Commons" project.
Indeed, the fundamental question is: Do we still want/need/have to implement
linear algebra within Commons, knowing that several alternative matrix algebra
Java projects have appeared (or evolved a lot) since the MATH-765 report
collected many issues of "Commons Math" package
{{org.apache.commons.math3.linear}} (a.o. the mentioned EJML)?
Depending on the answer, your task list would either grow too large
(implementing a full-fledged linear algebra library needs a plan, not just the
intention to "Port and redevelop the libraries of [the linear package]"),
or shrink to almost nothing since we'd just _use_ EJML. [And we might indeed
do that (details TBD) for providing the necessary functionality to advance the
tickets mentioned in the STATISTICS project].
In light of this, and unless [~ericbarnhill] has another take on it, I suggest
that you modify your proposal. A high priority task is to work towards the
first official release of "Commons Numbers". Would you consider the [more
concrete (but diverse) list of tasks that should be
resolved|https://issues.apache.org/jira/issues/?filter=12345517&jql=project%20%3D%20NUMBERS%20AND%20labels%20%3D%20gsoc2019%20AND%20text%20~%20numbers]
to that end?
was (Author: erans):
Hi [~Ting223].
Thanks a lot for your interest in this project.
We certainly need to tackle the linear algebra issue; however the "how" is not
clear to me; then, somehow, this JIRA ticket escaped my attention, and we don't
have enough time left to discuss/research the implications and their impact on
your GSoC proposal...
Also because some aspects will require more opinions, the discussion must be
brought to the "dev" mailing list of the "Commons" project.
Indeed, the fundamental question is: Do we still want/need/have to implement
linear algebra within Commons, knowing that several alternative matrix algebra
Java projects have appeared (or evolved a lot) since this report collected many
issues of "Commons Math" package {{org.apache.commons.math3.linear}} (a.o. the
mentioned EJML)?
Depending on the answer, your task list would either grow too large
(implementing a full-fledged linear algebra library needs a plan, not just the
intention to "Port and redevelop the libraries of [the linear package]"),
or shrink to almost nothing since we'd just _use_ EJML. [And we might indeed
do that (details TBD) for providing the necessary functionality to advance the
tickets mentioned in the STATISTICS project].
In light of this, and unless [~ericbarnhill] has another take on it, I suggest
that you modify your proposal. A high priority task is to work towards the
first official release of "Commons Numbers". Would you consider the [more
concrete (but diverse) list of tasks that should be
resolved|https://issues.apache.org/jira/issues/?filter=12345517&jql=project%20%3D%20NUMBERS%20AND%20labels%20%3D%20gsoc2019%20AND%20text%20~%20numbers]
to that end?
> Port commons-math.linear to commons-numbers.linear
> --------------------------------------------------
>
> Key: NUMBERS-98
> URL: https://issues.apache.org/jira/browse/NUMBERS-98
> Project: Commons Numbers
> Issue Type: Task
> Reporter: Eric Barnhill
> Priority: Major
> Labels: GSoC2019, gsoc2019
>
> Apache commons is one of the most widely used supplementary resources by Java
> programmers and its mathematical functions are widely used.
> Basic operations of linear algebra such as matrix factorization are not only
> in use in scientific and technical fields but widely in industry, and an
> accessible and standalone library for this functionality would have a wide
> potential audience.
> This ticket is in three parts:
> # Port the libraries from commons-math.linear into commons-numbers.linear,
> removing unnecessary layers of abstraction and creating a simple, intuitive,
> standalone library.
> # The developer should familiarize themselves with best bractices in linear
> algebra such as those in the EJML, and redevelop the linear library to
> contain best practice implementations.
> # The developer should work with developers of other math and statistics
> projects to integrate their work where it can benefit those projects.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
