I am taking the recommended curriculum of the Machine Learning Specialization. I’ve finished the fist part linear and logistic regression and have moved onto the second session Advanced Learning Algorithms and I feel like there has been a shift in problem solving which I somehow missed. We were solving for the vectorW which is used to predict values but in this course W appears to be some randomized values and the training is done via another method which I don’t believe I’ve seen. Don’t think I’ve skipped anything. Is there something you can direct me to?
1 Like
What is the name of the assignment you are working on, or the lecture you are watching?
I am about to take the quiz " Practice quiz: Neural network implementation in Python". I feel like there was no explanation of this new approach. Did I miss a couple of lectures that address this?
This course introduces the use of TensorFlow and Numpy. It’s covered in the first few sets of lectures in Week 1.
TensorFlow handles training automatically - you just design the network layers, and the backpropagation process is built-into the layer definitions.
Neural Networks use random initialization of the weight values. This is critical when the model includes hidden layers.
Andrew discusses this somewhere in the course, I do not recall at the moment exactly where.