| clear query|facets|time |
Search criteria: position.
Results from 61 to 68 from
68 (3.822s).
|
|
|
Loading phrases to help you
refine your search...
|
|
Re: Fwd: weighted score - Mahout - [mail # user]
|
|
... is expressed in terms of a quadratic loss function. This gives you three possibilities, one where positive weights dominate, one where negative weights dominate and a third where they balance out...
|
|
|
Author: Ted Dunning,
2010-02-23, 23:07
|
|
|
Re: LLR Scoring question - Mahout - [mail # user]
|
|
... of the LLR score: rootLLR = signum(k11/k1* - k21/k2*) * sqrt(LLR) This score has two advantages over the basic LLR: a) it is positive where k11 is bigger than expected, negative where...
|
|
|
Author: Ted Dunning,
2010-01-12, 19:20
|
|
|
Re: Modelling typed vectors? - Mahout - [mail # user]
|
|
... projections should be entirely composed of positive values. A better way to do this would be to use a log-linear (soft-max) link function to project R^n into the positive orthant. ...
|
|
|
Author: Ted Dunning,
2010-10-13, 06:07
|
|
|
Fwd: default tree builder - Mahout - [mail # dev]
|
|
... positive results with bagging of DefaultTreeBuilder-id3 like implementation. However, there is seems to be an issue with building possible running into infinite recursion, i.e. stack overflow...
|
|
|
Author: Ted Dunning,
2010-10-06, 21:22
|
|
|
Re: Adding weighting to boolean data - Mahout - [mail # user]
|
|
... to be a large fraction of your data volume in order to have a distinctly positive effect. Similarly, it is helpful to introduce alternative recommendations in an A/B test fashion so that you get...
|
|
|
Author: Ted Dunning,
2010-07-22, 15:04
|
|
|
Fwd: weighted score - Mahout - [mail # user]
|
|
...Weights can't be negative and still be weights. You can have large (positive) weights on negative training examples (aka "not like this"), but you can't really have a negative weight...
|
|
... effectively than the positive examples attract it by using a looser loss function on negative examples. This also would take into account the fact that strong negative ratings are actually more like...
|
[+ show more]
[- hide]
| ... the right answer than the average case at large. Averaging is just such a case since the mean is just the least squares solution for positive weights. When you have negative weights that represent... |
|
|
Author: Ted Dunning,
2010-02-23, 19:21
|
|
|
Re: Understanding the SVD recommender - Mahout - [mail # user]
|
|
.... Note that since this is an SVD, S is diagonal and all elements are real (and positive, actually). Thus B* = B. Ish. But not customary. The old LSI article makes a better case than I can...
|
|
|
Author: Ted Dunning,
2010-06-04, 07:53
|
|
|
Re: Clustering techniques, tips and tricks - Mahout - [mail # user]
|
|
... a more elaborate model than identical Gaussians, the use of maximum likelihood instantly causes horrible problems because if you allow the variance to vary, the algorithm will position some...
|
|
|
Author: Ted Dunning,
2010-01-03, 18:26
|
|
|
|
