Learning Strategy for Top Down Approach

Part 1 (2020)
1

Typically multiple passes are required when you are doing a Top-Down approach. And with each pass, we will understand a little bit more. For example, first pass we are at 10,000 feet view, with the second pass we are at 5,000 feet view and with the third pass we are at the ground level.

how-to-do-fastai-lesson
how-to-do-fastai-lesson1272×660 203 KB

Pass 1

Most of us would be at this level.

  • Watch the lecture once for high-level ideas

  • Read the chapter corresponding to the book

  • Capture the high-level ideas & concepts (Big Picture)

  • Review the questionnaire provided and try attempting to answer based on your understanding (It’s okay if you cannot answer all.)

Engage with lesson material

  • Read and Run the cells in the notebook. Add comments to the cell and explain what the cell does in your own words. If you do not understand, add a TODO on it.

  • use doc(THING_I_WANT_TO_KNOW_MORE) eg: doc(untar_data), ?? untar_data

  • Create the notes and an outline [1]

  • Complete the HomeWork specific to Lesson (For eg Lesson 1)

  • read the chapter corresponding to the lesson

Pass 2 - Detailed Engagement

  • Re-run the notebook but re-write every line one by one

how-to-use-provided-notebooks
how-to-use-provided-notebooks1920×1080 189 KB

  • Execute each one and explain what the code cell is doing & why it is doing that in your own words.

  • Watch the lecture again for more detail

  • Try to complete the questionnaire this time if you cannot answer all the questions at the previous level

Pass 3 - More Detailed Engagement

  • Create the notebook but use a different dataset (Refer this for how to find different datasets)

  • Watch the lecture again more slowly

  • Read the Suggested Readings and summarize it in your own words

  • Present & Teach the lecture in your study group

Footnotes

[1] Notes Template to get you started

Template

Next Steps

  • [ACTIONS THAT YOU NEED TO TAKE DEPENDING ON THE PASS]

  • [eg: Setup paperspace]

Outline/Topics Covered

  • Eg: Top-Down Approach

KEYPOINTS

  • Eg: What is deep learning?

ADVICE

  • [Things that Jeremey say to do goes here]

  • [Gems that you found as advice in the forum]

TODO

  • [As you work through the notebook, add the TODO here to come back & revisit]

  • Items from here go to the NEXT STEPS.

QUESTIONS

  • [questions that you need to do more research or want to search/ask in forums]

  • [errors that you are stuck or blocked on]. Refer [2] , Search on the forums for 15 minutes then ask help in the forum

[2] How to debug your code and ask for help

References

14 Likes
2

Great approach! :slight_smile:

2 Likes