I would say that a real user gets a bigger vote relative to theoretical
complainers.
Simple at first is fine by me.
On Tue, Jun 9, 2009 at 4:09 PM, Jeff Eastman wrote:
> IIRC, the patch in M-65 works but was judged to be inadequate so I never
> committed it. After several subsequent postings on
I had a nice Skype chat with Sebastien last week and he has my
wholehearted support on these improvements. The sampleFromPosterior
implementation was a total naive hack which was actually an improvement
over always sampling from the prior. The fact that the current
implementation converges so n
IIRC, the patch in M-65 works but was judged to be inadequate so I never
committed it. After several subsequent postings on the requirements I
started another version but it ran into impossible
serialization/deserialization issues. Those could probably be addressed
now with Json but the patch i
On Jun 9, 2009, at 5:49 PM, Grant Ingersoll wrote:
I'm looking into the whole labels thing as well as Vector stuff and
I'm confused by a couple of things.
1. DirchletMapper assumes DenseVector implementation, no? Line 45?
2. Shouldn't DenseVector implement equals like SparseVector does?
FWIW, I'm happy w/ a simple solution right now, which may very well be
Jeff's initial patch. Still, I'd like to hear more from Ted, Jeff and
Karl.
On Jun 9, 2009, at 6:20 PM, Benson Margulies wrote:
OK, I came in to the middle and misunderstood.
This doesn't precisely seem to leave much o
OK, I came in to the middle and misunderstood.
This doesn't precisely seem to leave much of a space for another
person to join in, but I'd be happy to be corrected by some
combination of the people cited below.
On Tue, Jun 9, 2009 at 6:18 PM, Grant Ingersoll wrote:
>
> On Jun 9, 2009, at 6:11 PM
On Jun 9, 2009, at 6:11 PM, Benson Margulies wrote:
Grant,
AFAIK, there's a perfectly adequate version sitting out there in
MAHOUT-65, waiting for a committer to commit it.
If that's wrong and there's a concrete coding task I could undertake
that would render it committable, I'd be game.
T
Grant,
AFAIK, there's a perfectly adequate version sitting out there in
MAHOUT-65, waiting for a committer to commit it.
If that's wrong and there's a concrete coding task I could undertake
that would render it committable, I'd be game.
--benson
On Tue, Jun 9, 2009 at 6:02 PM, Grant Ingersoll
On Jun 9, 2009, at 5:44 PM, Benson Margulies wrote:
I've basically bailed on doing much with Mahout until something like
this commits.
The only way it ever gets better is by people pitching in, but, heh,
you know that already! ;-)
On Tue, Jun 9, 2009 at 2:57 PM, Grant Ingersoll
wrot
I'm looking into the whole labels thing as well as Vector stuff and
I'm confused by a couple of things.
1. DirchletMapper assumes DenseVector implementation, no? Line 45?
2. Shouldn't DenseVector implement equals like SparseVector does?
3. VectorView doesn't appear to implement asFormatStrin
[
https://issues.apache.org/jira/browse/MAHOUT-121?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12717832#action_12717832
]
Benson Margulies commented on MAHOUT-121:
-
I could make you a fast sparse vector, b
I've basically bailed on doing much with Mahout until something like
this commits.
On Tue, Jun 9, 2009 at 2:57 PM, Grant Ingersoll wrote:
> What are people doing for keeping track of their vectors and matrices?
> MAHOUT-65 attempted to address that, but it seems to have gotten stuck.
>
> For MAHO
Fabulous!
Some details in-line.
On Mon, Jun 8, 2009 at 7:06 AM, Sebastien Bratieres wrote:
> - most importantly, the parameter re-estimation step currently is a maximum
> likelihood re-estimation. So the algorithm is not guaranteed to do actual
> training/converge/work at all. In order to do Gi
[
https://issues.apache.org/jira/browse/MAHOUT-121?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12717812#action_12717812
]
Grant Ingersoll commented on MAHOUT-121:
I've been doing some profiling and we do i
Dear Mahout developers,
I am planning to contribute to the Dirichlet Process Clustering algorithm
implemented by Jeff (Eastman). I have read through the code in some detail,
and discussed a couple of points with Jeff already in order not to create a
mess. That way I could understand how the code o
What are people doing for keeping track of their vectors and
matrices? MAHOUT-65 attempted to address that, but it seems to have
gotten stuck.
For MAHOUT-126 (document clustering prep work), I ended up outputting
Vector cell information to a separate file, which works but is
cumbersome.
[
https://issues.apache.org/jira/browse/MAHOUT-126?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12717742#action_12717742
]
Grant Ingersoll commented on MAHOUT-126:
Note, I haven't actually tried clustering
[
https://issues.apache.org/jira/browse/MAHOUT-126?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Grant Ingersoll updated MAHOUT-126:
---
Attachment: MAHOUT-126.patch
Here's a first attempt at my thoughts based on the two previous
[
https://issues.apache.org/jira/browse/MAHOUT-130?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Grant Ingersoll resolved MAHOUT-130.
Resolution: Fixed
Committed Ted's patch
> Vector should allow for other normalize powers t
[
https://issues.apache.org/jira/browse/MAHOUT-130?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12717594#action_12717594
]
Sean Owen commented on MAHOUT-130:
--
Regarding IntelliJ warnings -- double-check that your
20 matches
Mail list logo