[ https://issues.apache.org/jira/browse/SPARK-3181?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17148452#comment-17148452 ]
zhengruifeng commented on SPARK-3181: ------------------------------------- I am working on blockify+gemv/gemm for better performance, and found that the huber regression' s solution varies significantly when blockSize is changed. Suggested by [~weichenxu123], I tested the stability by [shuffling the input dataset|https://issues.apache.org/jira/browse/SPARK-32060?focusedCommentId=17148412&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-17148412], and found that: the solution of LInearSVC and LinearRegression(SquaredError) looks stable when input is shuffled; while the solution of LinearRegression(huber) varies significantly; I also tested scikit-learn's {{HuberRegressor which also use LBFGSB}} as the solver, its solutions seem stable. So I guess there maybe something wrong in {{BreezeLBFGSB}}. Maybe using 'Replace {{\sigma}} to {{\exp(\alpha)' or 'Pseudo-Huber loss' is an alternative.}} > Add Robust Regression Algorithm with Huber Estimator > ---------------------------------------------------- > > Key: SPARK-3181 > URL: https://issues.apache.org/jira/browse/SPARK-3181 > Project: Spark > Issue Type: New Feature > Components: ML > Affects Versions: 2.2.0 > Reporter: Fan Jiang > Assignee: Yanbo Liang > Priority: Major > Labels: features > Fix For: 2.3.0 > > Original Estimate: 0h > Remaining Estimate: 0h > > Linear least square estimates assume the error has normal distribution and > can behave badly when the errors are heavy-tailed. In practical we get > various types of data. We need to include Robust Regression to employ a > fitting criterion that is not as vulnerable as least square. > In 1973, Huber introduced M-estimation for regression which stands for > "maximum likelihood type". The method is resistant to outliers in the > response variable and has been widely used. > The new feature for MLlib will contain 3 new files > /main/scala/org/apache/spark/mllib/regression/RobustRegression.scala > /test/scala/org/apache/spark/mllib/regression/RobustRegressionSuite.scala > /main/scala/org/apache/spark/examples/mllib/HuberRobustRegression.scala > and one new class HuberRobustGradient in > /main/scala/org/apache/spark/mllib/optimization/Gradient.scala -- This message was sent by Atlassian Jira (v8.3.4#803005) --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org