systemml-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Deron Eriksson <deroneriks...@gmail.com>
Subject Re: Matrix Market format with metadata file
Date Tue, 16 Feb 2016 00:30:11 GMT
Thank you, Shirish. That makes sense. I'll update the docs to include this
information.

Deron


On Mon, Feb 15, 2016 at 4:26 PM, Shirish Tatikonda <
shirish.tatikonda@gmail.com> wrote:

> Both "mm" and "text" formats are identical except for a couple of
> differences:
>
> 1) for "mm": the matrix metadata is included in the first two lines; and
> for "text": the metadata is present in the associated .mtd file
> 2) "mm" data must be in a single file (i.e., no *part* files) where "text"
> data can span multiple *part* files (like any other file on HDFS).
>
> The support for "mm" is created mainly for the purpose of
> importing/exporting data in the format that R likes.
>
> Shirish
>
> On Mon, Feb 15, 2016 at 4:17 PM, Deron Eriksson <deroneriksson@gmail.com>
> wrote:
>
> > Hi,
> >
> > I have a question with regards to text vs mm. Isn't the mm coordinate
> > format identical to the text format but the mm data file happens to
> include
> > the metadata line for rows, cols, and nnzs, so shouldn't they scale the
> > same since the text row values (i,j,v) correspond to the mm rows?
> >
> > If we have the following MM:
> > %%MatrixMarket matrix coordinate real general
> > 4 3 6
> > 1 1 1.0
> > 1 2 2.0
> > 1 3 3.0
> > 3 1 7.0
> > 3 2 8.0
> > 3 3 9.0
> >
> > The corresponding text format (with accompanying metadata file) is:
> > 1 1 1.0
> > 1 2 2.0
> > 1 3 3.0
> > 3 1 7.0
> > 3 2 8.0
> > 3 3 9.0
> >
> > So aren't these formats essentially the same?
> >
> > Deron
> >
> >
> > On Mon, Feb 15, 2016 at 3:56 PM, Matthias Boehm <mboehm@us.ibm.com>
> wrote:
> >
> > > The meta data file is still useful in order to get the format. In case
> of
> > > matrix market, errors will be raised if included meta data is
> > inconsistent.
> > > So no, we should not disallow to specify the meta data. In general, we
> > > anyway recommend using text (textcell) instead mm (matrix market) for
> > > scalability reasons.
> > >
> > > Regards,
> > > Matthias
> > >
> > > [image: Inactive hide details for Deron Eriksson ---02/15/2016 03:45:46
> > > PM---Hi, The Matrix Market coordinate format contains # rows, #]Deron
> > > Eriksson ---02/15/2016 03:45:46 PM---Hi, The Matrix Market coordinate
> > > format contains # rows, # columns, and #
> > >
> > > From: Deron Eriksson <deroneriksson@gmail.com>
> > > To: dev@systemml.incubator.apache.org
> > > Date: 02/15/2016 03:45 PM
> > > Subject: Matrix Market format with metadata file
> > > ------------------------------
> > >
> > >
> > >
> > > Hi,
> > >
> > > The Matrix Market coordinate format contains # rows, # columns, and #
> > > non-zero values as metadata near the top of a matrix data file.
> > >
> > > If I write a matrix in mm format using SystemML, no metadata file is
> > > created since the metadata is stored within the data file.
> > >
> > > However, when reading a matrix with mm format, I can supply a metadata
> > > file, even though metadata exists in the matrix data file. Is there any
> > > reason for this, or should this be disallowed since the metadata file
> is
> > > redundant and can cause confusion, since metadata values can then be
> > > specified in two places, which then brings up the question, "which
> > metadata
> > > value should be used"?
> > >
> > > Deron
> > >
> > >
> > >
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message