diff options
author | Paul Mineiro <paul-github@mineiro.com> | 2014-01-02 05:47:38 +0400 |
---|---|---|
committer | Paul Mineiro <paul-github@mineiro.com> | 2014-01-02 05:47:38 +0400 |
commit | 89540c155e272333edc45ceb24e98e7295c4e00c (patch) | |
tree | c07bedd121fd6e331cb80223641a560aad68133f /demo/movielens | |
parent | 5f8f19adb47946cb2bf6f41962131be73a9ed878 (diff) |
tweak movielens demo
Diffstat (limited to 'demo/movielens')
-rw-r--r-- | demo/movielens/Makefile | 4 | ||||
-rwxr-xr-x | demo/movielens/README.md | 2 |
2 files changed, 3 insertions, 3 deletions
diff --git a/demo/movielens/Makefile b/demo/movielens/Makefile index f6ef4411..ece8d9e1 100644 --- a/demo/movielens/Makefile +++ b/demo/movielens/Makefile @@ -95,9 +95,9 @@ lrqdropout.results: ml-1m.ratings.test.vw ml-1m.ratings.train.vw @echo "* vw --lrq um12 --lrqdropout ... *" @echo "******************************************************" @echo - @${VW} --loss_function quantile -l 1 -b 24 --passes 100 \ + @${VW} --loss_function quantile -l 0.45 -b 24 --passes 100 \ -k --cache_file $@.cache -d $(word 2,$+) --holdout_off \ - --lrq um12 --lrqdropout --adaptive --invariant -f $@.model + --lrq um14 --lrqdropout --adaptive --invariant -f $@.model @echo "*****************************************************" @echo "* testing low-rank interaction model (with dropout) *" @echo "*****************************************************" diff --git a/demo/movielens/README.md b/demo/movielens/README.md index 4b6a49ab..674ca02b 100755 --- a/demo/movielens/README.md +++ b/demo/movielens/README.md @@ -35,7 +35,7 @@ You might find a bit of `--l2` regularization improves generalization. - `make shootout`: eventually produces three results indicating test MAE (mean absolute error) on movielens-1M for - linear: a model without any interactions. basically this creates a user bias and item bias fit. this is a surprisingly strong baseline in terms of MAE, but is useless for recommendation as it induces the same item ranking for all users. It achieves test MAE of 0.731. - lrq: the linear model augmented with rank-7 interactions between users and movies, aka, "seven latent factors". It achieves test MAE of 0.698. I determined that 7 was the best number to use through experimentation. The additional `vw` command-line flags vs. the linear model are `--l2 1e-6 --lrq um7`. Performance is sensitive to the choice of `--l2` regularization strength. - - lrqdropout: the linear model augmented with rank-12 interactions between users and movies, and trained with dropout. It achieves test MAE of 0.689. The additional `vw` command-line flags vs. the linear model are `--lrq um12 --lrqdropout`. + - lrqdropout: the linear model augmented with rank-14 interactions between users and movies, and trained with dropout. It achieves test MAE of 0.688. The additional `vw` command-line flags vs. the linear model are `--lrq um14 --lrqdropout`. - the first time you invoke `make shootout` there is a lot of other output. invoking it a second time will allow you to just see the cached results. Details about how `vw` is invoked is in the `Makefile`. |