The most important thing I learned from fast.ai wasn’t about deep learning, it was about teaching.
It’s better to teach top down. Here’s how you do it:
Show an important real world problem
Share a practical and complete solution to that problem
Progressively dig deeper into that solution to refine the understanding
I give machine learning trainings to professionals. And I’ve found the top-down approach effective for this target audience. Here’s why:
If you teach top-down you deliver most of the value in the first hours of the training.
This is especially important for training managers.
These are the people who hire the trainers. They don’t want to spend a lot of money on a course where only a fraction of their colleagues will learn something that’s of value to their organization.
Imagine teaching bottom-up. If some learners get stuck on an exercise then you’re in trouble, because helping the slowest means you could risk all learners not achieving the learning goal.
Teaching top-down means that you get breathing room immediately after you get your learners across the first hurdle.
Also, because the learners who’ve already “gotten it” are more happy to help the others too. They’ve gotten most of the value of the training. So they’re not racing towards the end anymore.
You can help those who need it most without forgetting about the others.
Face it. Not everyone signs up for professional trainings because they want to. Some people get sent to these trainings.
Starting with practical examples is a great way to engage them.
Progressively harder questions, especially ones aimed at critiquing the techniques they’ve just been taught are a nice trick to keep them on board.
…And the try-hards 🤩? They can just go deeper and deeper. There’s no end.
React on 🐦twitter.