[
https://issues.apache.org/jira/browse/MAHOUT-1732?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Dmitriy Lyubimov updated MAHOUT-1732:
-------------------------------------
Description:
Right now, the serialization of tensor types (matrix, vector) is double-bridged
thru VectorWritable and MatrixWritable.
a new writeup of kryo serializers in this issue provides direct support for
serializing vector and matrix types (also accounting for their structural
flavors).
This is a first rewrite, so improvements probably could be provided, but i so
far have not found this to become a visible bottleneck again over other spark
io costs.
https://github.com/apache/mahout/pull/135
was:
Right now, the serialization of tensor types (matrix, vector) is double-bridged
thru VectorWritable and MatrixWritable.
a new writeup of kryo serializers in this issue provides direct support for
serializing vector and matrix types (also accounting for their structural
flavors).
This is a first rewrite, so improvements probably could be provided, but i so
far have not found this to become a visible bottleneck again over other spark
io costs.
> Native support for kryo serialization of tensor types.
> ------------------------------------------------------
>
> Key: MAHOUT-1732
> URL: https://issues.apache.org/jira/browse/MAHOUT-1732
> Project: Mahout
> Issue Type: Improvement
> Reporter: Dmitriy Lyubimov
> Assignee: Dmitriy Lyubimov
> Fix For: 0.10.2
>
>
> Right now, the serialization of tensor types (matrix, vector) is
> double-bridged thru VectorWritable and MatrixWritable.
> a new writeup of kryo serializers in this issue provides direct support for
> serializing vector and matrix types (also accounting for their structural
> flavors).
> This is a first rewrite, so improvements probably could be provided, but i so
> far have not found this to become a visible bottleneck again over other spark
> io costs.
> https://github.com/apache/mahout/pull/135
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
