I want to get my hands dirty with deep learning. Not just to play with it but to do research and contribute something really interesting to the community.
I’ve heard books are really good to understand the concepts and are better than MOOC s. I also found the same when I recently completed reading “100page Machine learning book” by Andriy Burkov and currently I am reading “Grokking Deep learning” by Andrew Trask.
So, I want to know your opinions on learning deep learning from MOOC s vs Books. Which one do you think is better for research perspective?
You might feel like you’ve learned a lot by reading a book, but you probably don’t really understand most of it until you put the knowledge into practice. That means doing actual projects. With a MOOC there is often more emphasis on doing projects, which helps you learn things more thoroughly.
Not that books are bad or anything, but you need to combine book-learning with doing projects (for example from a MOOC) to get the most out of it.
yeah, thanks a lot!
That is what I am planning to do too!
I personally have limited book experience. They tend to go into more of the history and why, and not so much recent advancements. I think moocs are better for hands-on and more current learning, as they tend to be updated regularly. Books tend to do better for teaching long running concepts.
The fastai book is nice because it is written beside the course, and the course is updated frequently.
Hi saiyan6174 I hope you are having a jolly day!
I agree with the points. made by machinethink and marii.
I would add that with fastai you have a book, some videos and an active MOOC with practical exercises.
If you can answer all the questions and further research in the book, convert each of the notebooks into a working app, in my opinion you will definitely achieve your objective of:
People have different learning styles so I would use either or both, which ever helps you reach your goal the quickest. Experiment!
I’d recommend one Deep Learning Book that I’ve read from cover to cover more than once
It’s the classic trade-off between active vs passive learning.
MOOCs like fast.ai are sort of active learning as you play around with notebooks, get stuck, do exercises and apply the ideas to your own projects.
With books, you will need to design something to apply what you are reading. One idea could be to summarize what you learned at end of each chapter in a blog post or a notebook. I personally do this for research papers.
Alternatively, for every chapter you learn, try to implement it in code for some toy project.