Sorry to say but Course2 week 1 things do not seem to be in order. Not as systematic as things were in Course1. Spending more time to connect theory with Lab and for Lab when we have a new topic like tensorflow I feel it had to be more detailed explanations with more examples…Please if you could share some extra examples to for better understanding of tensorflow.
Now in Lab 03 the idea is to complete the coffee roasting exercise using numpy…but just after defining the functions, you straightway copy trained weights and biases from the previous lab in Tensorflow ? Also should the normalization also be done in numpy rather than using tensorflow function?
Why did we not do it in numpy in a complete manner? Is it available somewhere? I am planning to download the csv file and do it offline, but not sure if I will manage to get the same weights…
Normalization was covered in Course 1 in the video Feature scaling part 2 and implemented in Numpy in Optional lab: Feature engineering and Polynomial regression.
The documentation for the normalization function that the TensorFlow normalization layer uses was linked in Lab: Coffee Roasting in TensorFlow Normalization layer If you wish to reference it when implementing it yourself.
If you wish for more examples of TensorFlow being used I suggest their official tutorials
Yes the weights were used from the previous lab, the main objective of the lab is implementing the dense layer in numpy and showing that the implementation with the same weights as the TensorFlow version will produce the same output.
Thank you for your reply and sharing the links. This helps!!
However worried that without detailed tutorials on tensorflow which would help build our understanding, will we be able to implement / use the knowledge gained from this course.
I feel it would be have been great to have few more videos on tensorflow, as this is a specialization course and not just an overview.
Gaining confidence by taking a course is a different feeling. And that is what is the basic expectation the moment I see a course from Andrew NG.