How to finish FastAI course
- Start from Lesson 1, watch the video, finish the exercise to get a taste of the APIs.
- Continue with the rest of the lessons. Make sure you finish the questionnaire, you can leave the further research questions for the time being.
- If you are stuck, then try to find solutions but don’t stay there too long.
- Lesson which talks about Neural Net foundations is an important one, and the one where you build SGD and a few other models from scratch. Spend extra time on them, a week isn’t a bad idea.
- Run all the notebooks yourself to see the outputs and understand what it happening, if you don’t understand something then refer the various resources.
- If there are additional resources used in the lesson such as excel sheets, they will be available on the repo, always replicate them, excel is a great way to learn things from basics on small examples.
- Once you are in the middle of the course, take a gander of the things and write a couple of blog pieces to crystallise your understanding. They can be on any topics of your choice.
- It’s always a good idea to re run the notebooks and then run them again on a dataset of your choice and see if you can get the good levels of accuracy and if not then what are the hurdles.
- Finish all the lessons at least once. Finishing something entirely gives you a psychological advantage.
- Before wrapping up, go through all the questionnaires from all the lessons and find all the tech papers that were discussed in the lessons.
You are done for 2 weeks.
After 2 weeks of break, go back to the lessons, from 1 o 2, whatever you prefer and rewatch at 1.5x speed or whatever you prefer. This time take a stab at all the further research questions.
The things to focus in this iteration:
- Further research questions
- In every lesson there is a central idea, such as SGD from scratch, neural net from scratch, design your own grad etc, try to implement them without taking external help.
- Fill any gaps in understanding of the subject matter through primary and secondary research.
After or during the first iteration, if there is an idea forming on a particular problem statement that you want to solve then start working in that direction.
Whatever you have learned in these 2 iterations, apply them on your problem statement. Try to finish it fromend to end i.e after finishing it, deploy it or productify it and write an extensive blogpost about it.