Hi there,
As I’m getting deeper (heh) into working with Keras, I’m finding myself wanting to adopt some kind of coding style that allows me to more easily between objects-that-act-as-functions (such as layers), objects-that-act-as-placeholders, and actual-python-variables-with-values. I of course can just come up with an arbitrary one myself, but I was wondering:
– If anyone is feeling a similar frustration
– If anyone else has seen anyone try to solve this problem
– If anyone has started using any kind of convention themselves
– If the above indeed a useful taxonomy of the different ways variables can operate within a computational graph framework, or if I’m just generalizing from too little experience. (like, for example, perhaps it’s more useful to just make a distinction between placeholder-objects and having-value objects, since objects within the keras functional API can kinda just be thought of as functions)