Re: Add support of distributed matrices for OLS regression. (PR #3030 / IGNITE-5846)
The documentation is corect, all changes are added for internals in reality, without changes in API. 17 нояб. 2017 г. 1:35 AM пользователь "Denis Magda"написал: > Hi Alex, > > Impressive, thanks for sharing this with us. Btw, do we need to tweak the > existing OLS documentation to mention the contributed changes? > https://apacheignite.readme.io/docs/ols-multiple-linear-regression > > — > Denis > > > On Nov 16, 2017, at 5:55 AM, Alexey Zinoviev > wrote: > > > > Hello, > > > > I've added support of distributed OLS regression per > > IGNITE-5846 [1] and would want to push these to master. > > > > Changes impact "ml" module, more details of portion of code can be > > found in PR #3030 [2]. Yury Babak [3], Oleg Ignatenko[4] did review and > > sanity > > check. > > > > The main idea was to keep OLS Regression as algorithm and add support of > > many operations on SparseDistributedMatrix and > SparseBlockDistributedMatrix. > > > > To implement this functionality I added SparseBlockDistributedVector and > > SparseDistributedVector classes. Also a few operations on distributed > > vectors and matrices were implemented using power of affinity. > > > > The successful build could be found here [5] > > > > > > [1] https://issues.apache.org/jira/browse/IGNITE-5846 > > [2] https://github.com/apache/ignite/pull/3030 > > [3] https://github.com/ybabak > > [4] https://github.com/oignatenko > > [5] > > https://ci.ignite.apache.org/viewLog.html?buildId=945536; > buildTypeId=Ignite20Tests_IgniteMl=buildResultsDiv > > > > Sincerely yours, > > Alex > >
Re: Add support of distributed matrices for OLS regression. (PR #3030 / IGNITE-5846)
Hi Alex, Impressive, thanks for sharing this with us. Btw, do we need to tweak the existing OLS documentation to mention the contributed changes? https://apacheignite.readme.io/docs/ols-multiple-linear-regression — Denis > On Nov 16, 2017, at 5:55 AM, Alexey Zinovievwrote: > > Hello, > > I've added support of distributed OLS regression per > IGNITE-5846 [1] and would want to push these to master. > > Changes impact "ml" module, more details of portion of code can be > found in PR #3030 [2]. Yury Babak [3], Oleg Ignatenko[4] did review and > sanity > check. > > The main idea was to keep OLS Regression as algorithm and add support of > many operations on SparseDistributedMatrix and SparseBlockDistributedMatrix. > > To implement this functionality I added SparseBlockDistributedVector and > SparseDistributedVector classes. Also a few operations on distributed > vectors and matrices were implemented using power of affinity. > > The successful build could be found here [5] > > > [1] https://issues.apache.org/jira/browse/IGNITE-5846 > [2] https://github.com/apache/ignite/pull/3030 > [3] https://github.com/ybabak > [4] https://github.com/oignatenko > [5] > https://ci.ignite.apache.org/viewLog.html?buildId=945536=Ignite20Tests_IgniteMl=buildResultsDiv > > Sincerely yours, > Alex
Add support of distributed matrices for OLS regression. (PR #3030 / IGNITE-5846)
Hello, I've added support of distributed OLS regression per IGNITE-5846 [1] and would want to push these to master. Changes impact "ml" module, more details of portion of code can be found in PR #3030 [2]. Yury Babak [3], Oleg Ignatenko[4] did review and sanity check. The main idea was to keep OLS Regression as algorithm and add support of many operations on SparseDistributedMatrix and SparseBlockDistributedMatrix. To implement this functionality I added SparseBlockDistributedVector and SparseDistributedVector classes. Also a few operations on distributed vectors and matrices were implemented using power of affinity. The successful build could be found here [5] [1] https://issues.apache.org/jira/browse/IGNITE-5846 [2] https://github.com/apache/ignite/pull/3030 [3] https://github.com/ybabak [4] https://github.com/oignatenko [5] https://ci.ignite.apache.org/viewLog.html?buildId=945536=Ignite20Tests_IgniteMl=buildResultsDiv Sincerely yours, Alex