Best practices learning from the MOOC

Fastai’s teaching philosophy is to start at the top and delve down later. However, now that I’m on lesson 3 and I look at the later titles I’m uncertain if I should continue waiting to look into some of my questions or delve into them myself. Maybe its better explained with an example.

In Lesson 1 we cover image recognition with dogs and cats. But I would like to understand the workings of the library more. For example, how do we add a third classification of “unknown” and dump all the problem images there? I decided not to look into that expecting there would be more lectures down the road going into these details as per the top down teaching but the future lecture titles don’t hint at such a thing.

Obviously the MOOC is not meant to be a completely structured class but it seems natural to provide some guidance as to how deeply one should investigate matters before moving on.


I’m wondering about similar issues. In what detail should we look at the libraries? Are there visual representations of the libraries (pytorch/fastai) to sort of get a sense of what the overall structure is like or should we just read through the code to make sense of how things actually fit together.

The naming scheme is terse so for people like me who are not used to grokking library architectures on first reading (unfamiliarity with Python too) it is slow going.