[SPARK-6336] LBFGS should document what convergenceTol means

LBFGS uses convergence tolerance. This value should be written in document as an argument.

Author: lewuathe <[email protected]>

Closes #5033 from Lewuathe/SPARK-6336 and squashes the following commits:

e738b33 [lewuathe] Modify text to be more natural
ac03c3a [lewuathe] Modify documentations
6ccb304 [lewuathe] [SPARK-6336] LBFGS should document what convergenceTol means

Author: Lewuathe

docs/mllib-optimization.md

@@ -203,6 +203,10 @@ regularization, as well as L2 regularizer.
 recommended.
 * `maxNumIterations` is the maximal number of iterations that L-BFGS can be run.
 * `regParam` is the regularization parameter when using regularization.
+* `convergenceTol` controls how much relative change is still allowed when L-BFGS
+is considered to converge. This must be nonnegative. Lower values are less tolerant and
+therefore generally cause more iterations to be run. This value looks at both average
+improvement and the norm of gradient inside [Breeze LBFGS](https://github.com/scalanlp/breeze/blob/master/math/src/main/scala/breeze/optimize/LBFGS.scala).
 
 The `return` is a tuple containing two elements. The first element is a column matrix
 containing weights for every feature, and the second element is an array containing 

mllib/src/main/scala/org/apache/spark/mllib/optimization/LBFGS.scala

@@ -60,6 +60,8 @@ class LBFGS(private var gradient: Gradient, private var updater: Updater)
   /**
    * Set the convergence tolerance of iterations for L-BFGS. Default 1E-4.
    * Smaller value will lead to higher accuracy with the cost of more iterations.
+   * This value must be nonnegative. Lower convergence values are less tolerant
+   * and therefore generally cause more iterations to be run.
    */
   def setConvergenceTol(tolerance: Double): this.type = {
     this.convergenceTol = tolerance
@@ -142,7 +144,9 @@ object LBFGS extends Logging {
    *                   one single data example)
    * @param updater - Updater function to actually perform a gradient step in a given direction.
    * @param numCorrections - The number of corrections used in the L-BFGS update.
-   * @param convergenceTol - The convergence tolerance of iterations for L-BFGS
+   * @param convergenceTol - The convergence tolerance of iterations for L-BFGS which is must be
+   *                         nonnegative. Lower values are less tolerant and therefore generally
+   *                         cause more iterations to be run.
    * @param maxNumIterations - Maximal number of iterations that L-BFGS can be run.
    * @param regParam - Regularization parameter
    *