Repository: spark Updated Branches: refs/heads/master c2fe3a6ff -> b19605619
[SPARK-5537][MLlib][Docs] Add user guide for multinomial logistic regression Adding more description on top of #4861. Author: DB Tsai <[email protected]> Closes #4866 from dbtsai/doc and squashes the following commits: 37e9d07 [DB Tsai] doc Project: http://git-wip-us.apache.org/repos/asf/spark/repo Commit: http://git-wip-us.apache.org/repos/asf/spark/commit/b1960561 Tree: http://git-wip-us.apache.org/repos/asf/spark/tree/b1960561 Diff: http://git-wip-us.apache.org/repos/asf/spark/diff/b1960561 Branch: refs/heads/master Commit: b196056190c569505cc32669d1aec30ed9d70665 Parents: c2fe3a6 Author: DB Tsai <[email protected]> Authored: Mon Mar 2 22:37:12 2015 -0800 Committer: Xiangrui Meng <[email protected]> Committed: Mon Mar 2 22:37:12 2015 -0800 ---------------------------------------------------------------------- docs/mllib-linear-methods.md | 10 ++++++++++ 1 file changed, 10 insertions(+) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/spark/blob/b1960561/docs/mllib-linear-methods.md ---------------------------------------------------------------------- diff --git a/docs/mllib-linear-methods.md b/docs/mllib-linear-methods.md index 03f90d7..9270741 100644 --- a/docs/mllib-linear-methods.md +++ b/docs/mllib-linear-methods.md @@ -784,9 +784,19 @@ regularization parameter (`regParam`) along with various parameters associated w gradient descent (`stepSize`, `numIterations`, `miniBatchFraction`). For each of them, we support all three possible regularizations (none, L1 or L2). +For Logistic Regression, [L-BFGS](api/scala/index.html#org.apache.spark.mllib.optimization.LBFGS) +version is implemented under [LogisticRegressionWithLBFGS] +(api/scala/index.html#org.apache.spark.mllib.classification.LogisticRegressionWithLBFGS), and this +version supports both binary and multinomial Logistic Regression while SGD version only supports +binary Logistic Regression. However, L-BFGS version doesn't support L1 regularization but SGD one +supports L1 regularization. When L1 regularization is not required, L-BFGS version is strongly +recommended since it converges faster and more accurately compared to SGD by approximating the +inverse Hessian matrix using quasi-Newton method. + Algorithms are all implemented in Scala: * [SVMWithSGD](api/scala/index.html#org.apache.spark.mllib.classification.SVMWithSGD) +* [LogisticRegressionWithLBFGS](api/scala/index.html#org.apache.spark.mllib.classification.LogisticRegressionWithLBFGS) * [LogisticRegressionWithSGD](api/scala/index.html#org.apache.spark.mllib.classification.LogisticRegressionWithSGD) * [LinearRegressionWithSGD](api/scala/index.html#org.apache.spark.mllib.regression.LinearRegressionWithSGD) * [RidgeRegressionWithSGD](api/scala/index.html#org.apache.spark.mllib.regression.RidgeRegressionWithSGD) --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
