Vector.getDistanceSquared() is incorrect for both SparseVector varieties
Key: MAHOUT-268
URL: https://issues.apache.org/jira/browse/MAHOUT-268
Project: Mahout
Issue Typ
Vector.norm(x) uses incorrect formula for both x == POSITIVE_INFINITY and x == 1
Key: MAHOUT-267
URL: https://issues.apache.org/jira/browse/MAHOUT-267
Project: Mahout
[
https://issues.apache.org/jira/browse/MAHOUT-263?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jake Mannix updated MAHOUT-263:
---
Attachment: MAHOUT-263.diff
Let's try a patch which is svn up'ed first.
Suggest better names for t
[
https://issues.apache.org/jira/browse/MAHOUT-263?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jake Mannix updated MAHOUT-263:
---
Attachment: MAHOUT-263.diff
Ugly ugly names. Better suggestions?
> Matrix interface should extend I
[
https://issues.apache.org/jira/browse/MAHOUT-263?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jake Mannix updated MAHOUT-263:
---
Attachment: (was: MAHOUT-263.diff)
> Matrix interface should extend Iterable for better integrati
This is a fine way to go (and I have often (mis)used search engines as
recommendation engines).
Another angle is to consider the item level recommendations for a single
item to simply be additional attributes. You can also look at user level
cooccurrence analysis of attributes (including SVD) as
Greetings! Fellow GSOC alums, administrators and dear mentors, the next
edition is right here. Details are given in the link below.
https://groups.google.com/group/google-summer-of-code-discuss/browse_thread/thread/d839c0b02ac15b3f
Maybe we could identify key areas in Mahout which we need to deve
Yeah, this was one of my thoughts with MAHOUT-185 - turn some of our Driver
classes to just fire off a Tool. It is very convenient to be able to do
this, and
it's becoming more standard as well.
I need to dig up my stuff in decomposer/contrib-hadoop and pull that in and
integrate
it with Drew's p
Hi,
I'm using the PFPGrowth code (http://issues.apache.org/jira/browse/MAHOUT-157)
from Mahout 0.3 and it works fine on my local box. However, when I try to get
it to run on our grid cluster, it amazingly does not allow any parameters to be
passed to Hadoop. When I look at the code
(mahout/cor
Mahout algorithms are not using ToolRunner of Hadoop. I guess many core
hadoop-ers like that feature. I think we should be supporting that feature
by 0.3
Robin
On Wed, Jan 27, 2010 at 5:59 AM, Sean Owen wrote:
> These look like Hadoop params, to the hadoop command? why wouldn't
> hadoop be par
If you look at any of the Colt maps, you'll see a reference to double
hashing. Knuth's recommendation is to use twin primes. My meaning was
that the code doesn't have any twin primes, not that there are no such
beasts. I need to start at the code some more, but I was at least
briefly convinced that
On Tue, Jan 26, 2010 at 3:36 PM, Ted Dunning wrote:
> I define it a bit differently by redefining recommendations as machine
> learning.
>
> On Tue, Jan 26, 2010 at 1:44 PM, Sean Owen wrote:
>
> > I would narrow and specify this, in the context of Mahout, to have a
> > collaborative filtering an
On Tue, Jan 26, 2010 at 5:04 PM, Sean Owen wrote:
> You're saying content-based recommendation, in practice, is often a
> matter of substituting one dominant item attribute in place of items
> -- recommending on artist, rather than artist track. OK, check, one
> can do that in the current framewo
Nice, good wisdom here.
I agree about the appeal and problems of thinking of item-attribute
pairs as your items.
You're saying content-based recommendation, in practice, is often a
matter of substituting one dominant item attribute in place of items
-- recommending on artist, rather than artist
These look like Hadoop params, to the hadoop command? why wouldn't
hadoop be parsing those, or, why would the Job command have to shuttle
them to Hadoop? I thought these were typically set in the config .xml
files anyhow.
On Tue, Jan 26, 2010 at 11:43 PM, Aurora Skarra-Gallagher
wrote:
> Hi,
>
>
Code reference?
On Tue, Jan 26, 2010 at 11:42 PM, Benson Margulies
wrote:
> The double hashes in the Colt code don't look entirely, ahem,
> conventional to me.
>
> There aren't two different primes, let alone two different primes that
> differ by 2.
>
> I'm probably not reading carefully enough.
There are definitely twin primes that differ by 2 (29 and 31 for example).
See http://en.wikipedia.org/wiki/Twin_prime
Surely that isn't what you are talking about and my confusion results from
inability to correct a typo when reading.
On Tue, Jan 26, 2010 at 3:42 PM, Benson Margulies wrote:
> T
The double hashes in the Colt code don't look entirely, ahem,
conventional to me.
There aren't two different primes, let alone two different primes that
differ by 2.
I'm probably not reading carefully enough.
I define it a bit differently by redefining recommendations as machine
learning.
Users have preferences for objects with attributes.
We would like to learn from all user/object/attribute preference data to
predict so-far unobserved preferences of a user for other objects.
Normal recommendations
I want to knock down some support for content based recommendation.
And I want to solicit ideas about what this even means to its intended
audience -- users.
I define it broadly as a recommender in which:
- items have attributes (e.g. books have genres, titles, authors)
rather than being completel
Hi all,
I will do my best to get this in 0.3 release.
{quote}
> MAHOUT-232 Implementation of sequential SVM solver based on Pegasos
>
This patch looks to be progressing - it would be really nice to get it in.
{quote}
Cheers,
Zhendong
--
21 matches
Mail list logo
