network profile
| bio | website | |
|---|---|---|
| location | ||
| age | ||
| visits | member for | 1 year, 11 months |
| seen | Aug 12 '12 at 11:06 | |
| stats | profile views | 10 |
|
Aug 8 |
comment |
Initial values for logistic regression using maximum likelihood In addition, I've tried running glm, then taking the output coefficients and put them as start values to the mle. In this case I get almost the same results (and got acceleration in performance) |
|
Aug 8 |
comment |
Initial values for logistic regression using maximum likelihood for example, here are the first 7 coefficients in the glm run: -6.2307913208130525, 6.110187257533295e-06, -2.0577042478307273, 0.4786093240660332, 0.38126727104872804, -0.625615435816033, 0.04482479648912922 And the first 7 coefficients in the mle2 run (with zeros start values): -1.4625013759985311e-08, -0.0005046844214205488, 1.2295071793926704e-08, 1.632233091079531e-08, -1.8873962762517583e-08, -6.607669091467728e-09. You can see that it seems that it was "stuck" in a local minimum near 0,0,0,0,0,.... |
|
Aug 8 |
comment |
Initial values for logistic regression using maximum likelihood I'm not using glm, so as I understand safeBinaryRegression can't help me. What I've done is implemented my own log likelihood function and used maximum likelihood (R's mle2 - bbmle) to find the coefficients. I've compared the results to regular glm (which works without any problems). mle2 should get start values. Currently I'm using zeros for all the coefficients. When I'm using those start values, in some cases I get very weird results (the differences between the mle2 results and the glm results are big). So I wanted to know what are the start values that glm uses |
|
Aug 7 |
comment |
Initial values for logistic regression using maximum likelihood I have around 20,000 vectors with ~20 features per vector. The features has some dependencies between them. y values are 0 or 1 with around 1-5% or ones (hope that help, because I can't add all the data here... |
|
Aug 7 |
asked | Initial values for logistic regression using maximum likelihood |
|
Jun 14 |
comment |
Using generalized method of moments (GMM) to calculate logistic regression parameter Yes, we're looking for something that has probability between 0-0.25 and not between 0-1. |
|
Jun 14 |
comment |
Using generalized method of moments (GMM) to calculate logistic regression parameter We also trying to use Maximum likelihood, but would like to compare the results with GMM |
|
Jun 14 |
comment |
Using generalized method of moments (GMM) to calculate logistic regression parameter In the specific case where A=1 we get regular logistic regression |
|
Jun 14 |
awarded | Commentator |
|
Jun 14 |
comment |
Using generalized method of moments (GMM) to calculate logistic regression parameter fixed on all observations (like b0, b1, ...) |
|
Jun 14 |
comment |
Using generalized method of moments (GMM) to calculate logistic regression parameter either way. I can put it as an input (e.g.A = 0.25) or be one of the coefficients to be found |
|
Jun 14 |
asked | Using generalized method of moments (GMM) to calculate logistic regression parameter |
|
Dec 28 |
comment |
Changing logistic regression's loss function there are several different functions. I get the "score" for the probabilities estimation according to these functions, so I want, the fitted function to be estimated according to those "scores" (and not just by using maximum likelihood |
|
Dec 28 |
comment |
Changing logistic regression's loss function But is there any other method to fit a logit function to a 0s & 1s response using a different loss function? |
|
Dec 28 |
asked | Changing logistic regression's loss function |
|
Jul 22 |
awarded | Supporter |
|
Jul 22 |
comment |
Feature selection for low probability event prediction One more thing - I had some problems running "advanced" (RF, l1,...) classifiers on all my dataset (because of its size...), So I'm thinking of using weighting / data reduction, but again, I have a problem with finding the right probability (I found that finding the right probability after weighting is not straight forward..., though I haven't yet tried the methods that were mentioned above) Thanks! |
|
Jul 22 |
comment |
Feature selection for low probability event prediction I agree, however the problem with using l1/l2-regularized regression (LARS or logistic regressions), is that as far as I understand (and maybe I'm wrong), you don't get the right probability because of the regularization (is that correct?). In addition we found out that the logistic regression gives us better results in probability estimation than linear regression, when running on the entire dataset. |
|
Jul 21 |
comment |
Feature selection for low probability event prediction mmm... very interesting. In this case, do you think we should apply weighting (or "zero" reduction) before the classification? I will also check this. Thanks! |
|
Jul 21 |
comment |
Feature selection for low probability event prediction In addition, we can't weight our samples and still get the right probability. We haven't tried PCA yet, but its on our TODO :) Do you have any suggestions? Thanks a lot! |
