New to keras, tensorflow, layers - quick guide?

Course Q&A Deep Learning Specialization Convolutional Neural Networks
1

I’m a little confused as to how to proceed on Week 1, homework 2.
I incorrectly assumed we’d get some class presentation about how to use keras and tensorflow in this class and course.

Can you suggest any general resources that would help - other than the function definitions which is not quite general or broad enough, nor does it help with context.

Thanks a lot,
David

1 Like
2

Did you take Course 2 in this series? That’s where TF was introduced (C2 Week 3). In Course 4, they just give some examples of new constructs as they occur. If you want more background on the Keras Sequential and Functional APIs, here’s a thread from fellow student ai_curious which is really worth a look.

There are also lots of tutorials and examples on the TensorFlow website. The only thing you need to be a bit cautious about is that we are using a version of TF that is more than a year old and things mutate pretty fast. There are ways to request a particular version of any of the TF docs. You can find the version in any of our notebooks by doing:

print(tf.__version__)

3

Hi,
I did take course 2 (C2 w3) although, although I barely thought of that as an introduction. I kind of muddled my way through that. Actually having the instructor talk about it for a few minutes and giving some context, and explaining the toolset would be helpful at some point.
Anyway, I’ll check out the link you included, maybe watch a youtube video or two on the subject, and I’ll try to figure out this assignment.
Thanks, David

1 Like
4

One example of something that I find confusing, is that there is no input_shape listed as an available argument in ZeroPadding2D, yet if it’s not there the code fails, so I’m trying to understand where that information is located.
I went back to earlier versions of the tensorflow documentation and it’s not mentioned there either, including 2.8, 2.0, and 1.15.
Thanks, David

5

This is OOP, so they don’t document all the methods and attributes that any given subclass inherits from its parent classes. You may have to climb the inheritance tree to find things. All those classes (padding, pooling, conv, dense, yadda, yadda) inherit from Layer, so have a look there.

6

Thank you.
Great insight. I’ll take a look.
Best,
David