Google has brought deep learning a bit more mainstream. They also launched a short Udacity course, but it requires previous knowledge of Machine Learning. Which is another 10-12 week course on both Udacity and Coursera. I didn’t have that kind of spare time so I cheated. These are my notes on how I did it.
The reason these courses take so much time is because of all the theory and math to these algorithms. Which is good for writing accurate algorithms but a good hacker can just look it up when needed. If you, like me, just want to get your hands dirty asap then this post should help you wing it!
To fulfill prerequisites for the deep learning course, first we need to learn some ML algorithms and how to implement them in scikit-learn (a Python framework for Machine Learning).
The following plan should take a full weekend:
The first deep learning assignment is pretty hard so you are not alone. Here are some pointers for the final problem:
You already know a bit about linear regression, this assignment can be solved using an extended version of that, called Logistic Regression. You will need to read about it here. After that the sk-learn’s docs about it should make some sense. Give it a shot!
The assignment is also explained in great detail in the next chapter. So don’t worry if you don’t understand all the bits yet.
If you are successful in understanding and solving the first assignment, the rest of the course shouldn’t require any other resources.
I should emphasize that this isn’t a replacement for going through the full course. Choose this plan only if the alternative is not learning at all.
Great! Now you know deep learning, just don’t make another chat bot.
Anything but that...Good luck!