[ https://issues.apache.org/jira/browse/MADLIB-1286?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16710457#comment-16710457 ]
Rahul Iyer commented on MADLIB-1286: ------------------------------------ I don't think this is a bug - the test data does not need to be standardized since the coefficients returned are re-scaled back from the standard scale. This is possible since the βX product is linear. > Mean and std_dev from standardization not used in predict > --------------------------------------------------------- > > Key: MADLIB-1286 > URL: https://issues.apache.org/jira/browse/MADLIB-1286 > Project: Apache MADlib > Issue Type: Bug > Components: Module: Regularized Regression > Reporter: Nandish Jayaram > Priority: Major > Fix For: v1.16 > > > Elastic net train has a parameter called _standardize_. When this flag is > set, train creates a new temp table with standardized independent variable > and uses it for training. But, the mean and std_dev computed in this > standardization process is not captured in the output table. So during > predict, the independent variable is not standardized (with the mean and > std_dev that was used to standardize training data) if the model was trained > with _standardize=true_. > Is this expected behavior? If not, this seems like a bug. -- This message was sent by Atlassian JIRA (v7.6.3#76005)