From f87ae7177a5470a6a9f78bba9568317120ae02cc Mon Sep 17 00:00:00 2001 From: Xiangrui Meng Date: Wed, 11 Mar 2015 00:49:43 -0700 Subject: [PATCH 1/3] mention step size change --- docs/mllib-guide.md | 1 + 1 file changed, 1 insertion(+) diff --git a/docs/mllib-guide.md b/docs/mllib-guide.md index 4c7a7d9115ca1..64281a73a39aa 100644 --- a/docs/mllib-guide.md +++ b/docs/mllib-guide.md @@ -107,6 +107,7 @@ In the `spark.mllib` package, there were several breaking changes. The first ch * In `DecisionTree`, the deprecated class method `train` has been removed. (The object/static `train` methods remain.) * In `Strategy`, the `checkpointDir` parameter has been removed. Checkpointing is still supported, but the checkpoint directory must be set before calling tree and tree ensemble training. * `PythonMLlibAPI` (the interface between Scala/Java and Python for MLlib) was a public API but is now private, declared `private[python]`. This was never meant for external use. +* In linear regression (including Lasso and ridge regression), we scaled the squared loss by 0.5. So in order to produce the same result as in 1.2, the step size you chose needs to be scaled by 2. ## Previous Spark Versions From 375fd090850a48fa2d41f112103706da8d4a2127 Mon Sep 17 00:00:00 2001 From: Xiangrui Meng Date: Thu, 12 Mar 2015 01:47:15 -0700 Subject: [PATCH 2/3] address Sean's comments --- docs/mllib-guide.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/mllib-guide.md b/docs/mllib-guide.md index 64281a73a39aa..07f27e88f1c21 100644 --- a/docs/mllib-guide.md +++ b/docs/mllib-guide.md @@ -107,7 +107,7 @@ In the `spark.mllib` package, there were several breaking changes. The first ch * In `DecisionTree`, the deprecated class method `train` has been removed. (The object/static `train` methods remain.) * In `Strategy`, the `checkpointDir` parameter has been removed. Checkpointing is still supported, but the checkpoint directory must be set before calling tree and tree ensemble training. * `PythonMLlibAPI` (the interface between Scala/Java and Python for MLlib) was a public API but is now private, declared `private[python]`. This was never meant for external use. -* In linear regression (including Lasso and ridge regression), we scaled the squared loss by 0.5. So in order to produce the same result as in 1.2, the step size you chose needs to be scaled by 2. +* In linear regression (including Lasso and ridge regression), the squared loss is now divided by 2. So in order to produce the same result as in 1.2, the step size you choose needs to be multiplied by 2. ## Previous Spark Versions From fb3bbe6644549fb4ceb1d55024de30a5c18a79f7 Mon Sep 17 00:00:00 2001 From: Xiangrui Meng Date: Thu, 12 Mar 2015 16:04:52 -0700 Subject: [PATCH 3/3] mention regularization parameter --- docs/mllib-guide.md | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/docs/mllib-guide.md b/docs/mllib-guide.md index 50aec0b6178ff..f8e879496c135 100644 --- a/docs/mllib-guide.md +++ b/docs/mllib-guide.md @@ -102,7 +102,8 @@ In the `spark.mllib` package, there were several breaking changes. The first ch * In `DecisionTree`, the deprecated class method `train` has been removed. (The object/static `train` methods remain.) * In `Strategy`, the `checkpointDir` parameter has been removed. Checkpointing is still supported, but the checkpoint directory must be set before calling tree and tree ensemble training. * `PythonMLlibAPI` (the interface between Scala/Java and Python for MLlib) was a public API but is now private, declared `private[python]`. This was never meant for external use. -* In linear regression (including Lasso and ridge regression), the squared loss is now divided by 2. So in order to produce the same result as in 1.2, the step size you choose needs to be multiplied by 2. +* In linear regression (including Lasso and ridge regression), the squared loss is now divided by 2. + So in order to produce the same result as in 1.2, the regularization parameter needs to be divided by 2 and the step size needs to be multiplied by 2. ## Previous Spark Versions