How To Learn Deep Learning In Week or Months?

 

It's quite possible to learn, follow and contribute to state-of-art work in deep learning in about 6 months. This article details the way to achieve that.

Pre-requisites

-You're willing to spend 10 – 20 hours per week for the coming 6 months

-You have some programming skills. You should be comfortable picking up Python along the way. And cloud. (No background in Python and cloud assumed).

- Some math education in the past (algebra, figure, etc).

-Access to the internet and computer.

Step 1

We learn to drive a car — by driving. Not by learning how the clutch and the internal combustion engine work. At least not originally. When learning deep learning, we will follow the same top-down approach.

Do thefast.ai course — Practical Deep Learning for Coders — Part 1. This takes about 4 – 6 weeks of trouble. This course has a session on running the law on the cloud. Google Colaboratory has free GPU access. Start with that. Other options include Paperspace, AWS, GCP, Castle, and Floydhub. All of these are great. Don't start to make your machine. At least not yet.

Step 2

This is the time to know some of the basics. Learn about calculus and direct algebra.

For calculus, Big Picture of Calculus provides a good overview.

For Linear Algebra, Gilbert Strang’s MIT course on Open Courseware is amazing.

Once you finish the below two, read the Matrix Calculus for Deep Learning.

Step 3

Now is the time to understand the bottom-up approach to deep learning. Do all the 5 courses in the deep learning specialization in Coursera. You need to pay to get the assignments graded. But the trouble is truly worth it. Immaculately, given the background you have gained so far, you should be suitable to complete one course every week.

Step 4

“ All work and no play makes Jack a dull boy”

Do a capstone design. This is the time when you claw deep into a deep learning library (eg Tensorflow, PyTorch, MXNet) and apply an architecture from scratch for a problem.

The first three-way is about understanding how and where to use deep learning and gaining a solid foundation. This step is about implementing a project from scratch and developing a strong foundation on the tools.

Step 5

Now go and do fast.ai’s part II course — Cutting Edge Deep Learning for Coders. This covers more advanced topics and you'll learn to read the rearmost exploration papers and make sense out of them.

Each of the ways should take about 4 – 6 weeks. And in about 26 weeks since the time you started by CETPA Infotech.

Where to go next?

Do Stanford’s CS231n and CS224d courses. These two are amazing courses with great depth for vision and NLP independently. They cover the rearmost state-of-art. And read the deep learning book. This will solidify your understanding.