Perhaps the most well-known resource for learning deep learning is Andrew Ng’s series of 5 courses on Coursera. Those courses are still a great resource for anyone learning the fundamentals of the field but they are now a few years old (their launch was announced in August 2017). In this post, I will give you three main reasons why you should instead start from MIT’s course that I am going to tell you about. Read on for 3 reasons why you should start your journey with MIT’s Introductory Deep Learning Course:

Reason 1: The most up-to-date introductory course freely available right now

Yes, you will not find a more up-to-date course at the moment and that too, from one of the best institutions in the world for free.

Why is up-to-date important in deep learning?

The speed of progress in deep learning is phenomenal. To get an idea of the pace of progress, check this out!

The 2020 version of this course’s introduction (which was amazing in itself) contained Ex-President Barack Obama introducing the course (where the video was generated artificially). To create this video, a full actual video of Obama was required which was then picked and stitched together (believe it or not, the voice quality was deliberately degraded to make it obvious that this was not real).

However, in only 1 year, the field has advanced so rapidly and significantly that only a single static image is sufficient to create an artificial video of someone (you have to watch the first lecture of 2021 to see the demonstration! ).

Reason 2: The course has the right blend of breadth and depth for beginners

When you start learning about a new field, you are infinitely more motivated if you can get a sense of the big picture and a good feel of the breadth of the field in a short period. While you can spend your whole career within a small niche within deep learning, it would make sense if you, as a beginner, can get a decent breadth of the subject which is more than an hour’s lecture giving you just an overview but not so deep that you get lost in the nitty-gritty details of a single technique and lose sight of the overall picture. I think this course gets the balance right. It has both breadths to leave you with a very good understanding of the (current) scope of the field while going in sufficient detail to help you get cracking on with your own deep learning projects!

The course covers both the foundational theory (starting from the building blocks of deep learning: neural networks) and then moves on to cover well-known deep learning architectures and approaches (recurrent neural networks, convolutional neural networks, generative models, reinforcement learning, etc.), important issues when dealing with deep learning (ethical issues such as bias, fairness etc.) and exciting application areas (such as AI in health) via guest lecturers.

Reason 3: The pre-requisites for the course are minimal

Please don’t expect this course to be a walk in the park. You will have to put in the work and hours but the requirements to start making that effort are minimal. You need to have some knowledge of calculus (how to take derivatives), and linear algebra (how to multiply matrices). The programming exercises are based on Python (prior experience is not required but would, of course, be helpful).

To conclude, if you have not started learning about deep learning yet, I’d say start with MIT’s introductory course to get your feet wet. This post gave you three very good reasons for this choice (most up-to-date, the right blend of depth and breadth, and minimal pre-requisites). Once you are done with the MIT course, you should definitely check Andrew Ng’s 5-course sequence in Coursera which goes into more depth of the foundational concepts of the field.