1) new gradient descent notes 2) image of gradient descent in action reaching 2 different local minimum

Author: Quinn Liu

imagesForExplanation/GradientDescentWithMutlipleLocalMinimum.jpg

@@ -13,5 +13,4 @@ supervisedLearning
     - math notation = minimize over theta_0, theta_1 the cost function J(theta_0, theta_1) also called the squared error function
     - WHOLE POINT explained using a picture: https://github.com/quinnliu/MachineLearning/blob/master/imagesForExplanation/CostFunctionExampleWithTheta_0AndTheta_1.jpg 
     - Now plugging in the minimal theta_0 and theta_1 our function h(x) = theta_0 + theta_1 * x will predict h(x) = y by giving it an input x.
-
-    - GRADIENT DESCENT = algorithm that lets us find a minimal theta_0 and theta_1
+    - But how do we find the minimal theta_0 and theta_1?! GRADIENT DESCENT = algorithm that lets us find a minimal theta_0 and theta_1. It can also be used to minimize any arbitrary function J.