Discussion:
Matrix multiplication on Flink
Simone Robutti
2016-05-17 16:31:41 UTC
Permalink
Hello,

Suneel Marthi told me to engage the community about issues regarding matrix
multiplication on Mahout over Flink: I'm currently working on a
contribution to FlinkML that implements block matrix multiplication and it
seems to work but I still have to test it on a large scale. I would like to
know what issues were encountered in order to understand if my
implementation is affected.

Thanks,

Simone
Suneel Marthi
2016-05-17 16:36:31 UTC
Permalink
Glad to help and Welcome here. Maybe we could start with the implementation
that you have for MMul, if u could point us to that.

On Tue, May 17, 2016 at 12:31 PM, Simone Robutti <
Post by Simone Robutti
Hello,
Suneel Marthi told me to engage the community about issues regarding matrix
multiplication on Mahout over Flink: I'm currently working on a
contribution to FlinkML that implements block matrix multiplication and it
seems to work but I still have to test it on a large scale. I would like to
know what issues were encountered in order to understand if my
implementation is affected.
Thanks,
Simone
Simone Robutti
2016-05-17 16:49:46 UTC
Permalink
Here you should see the actual implementation.
https://github.com/radicalbit/flink/tree/FLINK-1873/flink-libraries/flink-ml/src/main/scala/org/apache/flink/ml/math/distributed

Mind that it's working but has never been reviewed and it's still a WIP.
Il 17/mag/2016 06:36 PM, "Suneel Marthi" <***@apache.org> ha scritto:

Glad to help and Welcome here. Maybe we could start with the implementation
that you have for MMul, if u could point us to that.

On Tue, May 17, 2016 at 12:31 PM, Simone Robutti <
Post by Simone Robutti
Hello,
Suneel Marthi told me to engage the community about issues regarding matrix
multiplication on Mahout over Flink: I'm currently working on a
contribution to FlinkML that implements block matrix multiplication and it
seems to work but I still have to test it on a large scale. I would like to
know what issues were encountered in order to understand if my
implementation is affected.
Thanks,
Simone
Loading...