[SPARK-16494][ML] Upgrade breeze version to 0.12

## What changes were proposed in this pull request?
breeze 0.12 has been released for more than half a year, and it brings lots of new features, performance improvement and bug fixes.
One of the biggest features is ```LBFGS-B``` which is an implementation of ```LBFGS``` with box constraints and much faster for some special case.
We would like to implement Huber loss function for ```LinearRegression``` ([SPARK-3181](https://issues.apache.org/jira/browse/SPARK-3181)) and it requires ```LBFGS-B``` as the optimization solver. So we should bump up the dependent breeze version to 0.12.
For more features, improvements and bug fixes of breeze 0.12, you can refer the following link:
https://groups.google.com/forum/#!topic/scala-breeze/nEeRi_DcY5c

## How was this patch tested?
No new tests, should pass the existing ones.

Author: Yanbo Liang <ybliang8@gmail.com>

Closes #14150 from yanboliang/spark-16494.

1 files changed, 5 insertions, 1 deletions

diff --git a/mllib/src/test/java/org/apache/spark/ml/feature/JavaPCASuite.java b/mllib/src/test/java/org/apache/spark/ml/feature/JavaPCASuite.java
index ac479c0841..8c0338e284 100644
--- a/mllib/src/test/java/org/apache/spark/ml/feature/JavaPCASuite.java
+++ b/mllib/src/test/java/org/apache/spark/ml/feature/JavaPCASuite.java
@@ -107,7 +107,11 @@ public class JavaPCASuite extends SharedSparkSession {
       .fit(df);
     List<Row> result = pca.transform(df).select("pca_features", "expected").toJavaRDD().collect();
     for (Row r : result) {
-      Assert.assertEquals(r.get(1), r.get(0));
+      Vector calculatedVector = (Vector) r.get(0);
+      Vector expectedVector = (Vector) r.get(1);
+      for (int i = 0; i < calculatedVector.size(); i++) {
+        Assert.assertEquals(calculatedVector.apply(i), expectedVector.apply(i), 1.0e-8);
+      }
     }
   }
 }