+1 Evolve from concrete examples look a practical path.
On 16 April 2012 08:39, Edward J. Yoon <[email protected]> wrote: > Let's implement above BSP applications in examples package first. > > If we can extract reusable code from these applications such as data > structure classes and important math algorithms, it should be added to > Hama core as you said. > > On Mon, Apr 16, 2012 at 1:26 AM, Thomas Jungblut > <[email protected]> wrote: >> Since we have in our latest release a graph package, we can focus a bit >> more on math. >> Should we therefore add a math package? >> >> I see several features to add: >> - Implementation of cholesky decomposition [1] >> - Implement the incomplete Cholesky Factorization [2] >> - Sparse Matrix-Vector multiplication [3] >> - A BSP matrix multiplication implementation [4] >> - Random matrix generator [5] >> - Matrix inversion [6] >> - SVD / Singular value decomposition [7] should we reopen this? >> etc. >> >> Should we create an umbrella for them and add a math package (mvn project)? >> >> Then what math lib should we use? (mahout-math, I have scripted one, a new >> one build by us). >> And yes we really need one, it makes things much simpler. >> >> How do we deal with examples? >> IMHO we should add the algorithms to the math package and add appropriate >> real-world use cases of them to the example package. For the algorithm use, >> we should document what input and output is to expect on a wiki page. >> >> >> [1] https://issues.apache.org/jira/browse/HAMA-94 >> [2] https://issues.apache.org/jira/browse/HAMA-553 >> [3] https://issues.apache.org/jira/browse/HAMA-524 >> [4] https://issues.apache.org/jira/browse/HAMA-221 >> [5] https://issues.apache.org/jira/browse/HAMA-500 >> [6] https://issues.apache.org/jira/browse/HAMA-220 >> [7] https://issues.apache.org/jira/browse/HAMA-176 >> >> -- >> Thomas Jungblut >> Berlin <[email protected]> > > > > -- > Best Regards, Edward J. Yoon > @eddieyoon
