How I Finished Andrew Ng’s Deep Learning Specialization in Just 4 Weeks

The secret of learning data science in the most efficient and fastest way

Andrew Ng is one of the famous heroes in the Artificial Intelligence field and his deep learning specialization on Coursera is one of the most leading courses in data science. I don’t doubt you have heard of it at least once. It is consist of 5 sub-courses covering from basic of neural networks to convolutional neural networks, to sequence models. Each course has 3 or 4 weeks of materials including the quiz and the code assignments (which was the greatest part of this course for me). So in total, this specialization requires approximately 3 months with 75 hours of materials to complete, and I finished it in 3 weeks and spent an additional 1 week to review the whole courses.

So I’m going to share how I could finish all these in such a short time. But before that, I feel the need to clarify that being fast is not exactly what I’m about to say. Everyone has his or her own pace of learning and having a solid understanding is much more important than being just a sprinter. ‘Three weeks’ is just about the number showing how much efficiency I could make. Additionally, this won’t be about a method like Pomodoro technique or a message like ‘just cut off all your distractions and turn off your phone.’ Cause that won’t work for me either. Then what was my secret? How did I make it?

But wait. Just for a second. Could I praise myself a little bit?

YoooOOOOO!!!! Whoho! Way To Gooooooooo!! Yeeeeeaah~🙌🙌🙌

Okay..

I think it’s quite enough for an introduction. Now here goes the main part.

The principle of learning new things

For just a little bit of my background, I majored material science and engineering at the undergraduate school. And then I worked for a battery company for a few years. Now I switched my career and I’m studying data science and business strategies for my Master’s degree. So.. Through the experience of learning all of these. Also through the experience of learning both R and Python so far. What I realized about ‘learning a new thing’ is.. It’s really a matter of familiarity.

I want to explain this concept with resistance analogy. When we are to drag or move something, there is a force resisting the relative motion on surfaces. One interesting feature is that the power of this force drops drastically after some amount of time. And from this point, the work becomes much easier than before. In physics, this is called ‘maximum static friction.’

And I believe this also belongs to learning. Not just about data science. Most of the cases in the world. If you are to study something with little background knowledge, you probably have to spend some amount of time filling with nothing but struggling. It is sometimes so harsh that it feels like you’re never going to make it. But if you endeavor and keep pushing ahead, then you finally pass a certain point when things become suddenly manageable or even more comfortable to go further. The phase when you finally get familiar with it. The stage where you finally passed ‘the maximum static friction’ point.

You’ll agree with me that there is a significant difference between learning something that you have heard of at least one time and learning something that you don’t. This is due to familiarity. Moreover, for most of the cases, there are some principles that repeat and overlap continuously within a subject. So when you understand these common basics, you can learn other modified or advanced concepts faster than before. Therefore when you are to achieve a new skill, one factor we have to fight against is being used to it. Then how could we get the familiarity in a faster and more efficient speed?

Read broadly like a hungry man

I was hungry. I was like a hungry man who never can feel the fullness in his stomach. So I read and read and read. I read articles on data science and machine learning from Toward Data Science, KDnuggets, Analytics Vidhya, and various email lists. That was because I had no idea with this field but was thirsty for the knowledge so badly. But there was no one who can tell this around me. The way for filling up this desire was reading. Cause there are tons of articles, which are free.

Of course, I couldn’t understand them 100% at first. Even though some of them were tutorials for the beginners, I was able to digest just 60% of them. It was stressful honestly. But I didn’t put too much effort for one single article. Instead, I read many articles about one topic and collected in my scraping sheet. As the number of the reading list increased one by one, my understanding also got better continuously.

I think these backgrounds made the magic. Even if I didn’t understand those resources thoroughly enough at those time, they have been piled on somewhere in my brain. And when they met Andrew Ng’s clear and intuitive instruction, all those ‘un-internalized’ knowledge finally could gather into one. So the course wasn’t that hard for me, and I felt a bit familiar with most of the materials already. It felt like I’d just passed the ‘maximum static friction’ point.

So I recommend you to read a lot and broadly. It’s okay you don’t understand them all at once. Just move on. It’s important not to be obsessed with them. The ideal way is reading at least more than 3 articles from different people and practice it on your own when you learn a new concept or an algorithm. This method could seem indirect and contouring. But I believe this is a process of constructing a robust baseline.

Don’t stick with just one Mooc

I also want to tell you that use various Moocs and blogs altogether. We have tons of tutorials and online courses: Datacamp, Udemy, Coursera, Edx, Kaggle’s learn section and Youtube. They are all good and they are all different. They have their own strength. For one instance, Datacamp, which I’ve been recommending several times, is great to focus on the essence of machine learning. Whereas Coursera, which is basically instructed by university institutes, have a little bit more academical approach and require more in-depth knowledge. Therefore don’t just stick one platform, and switch around during your study.

Moreover, KDnuggets, Analytics Vidhya and Kaggle Learn, the sites I mentioned above, can be another kind of ‘Mooc.’ Gratefully, there are lots of people who are willing to share their knowledge and skills through writings and videos. They usually deliver more details than regular courses. As online courses have to cover quite a lot of topics, we sometimes can find more detailed or practical information from blogs. So please enjoy and share those fruits. It could seem more time consuming at first, but trust me. They will pay you back in some days.

Last but not least

To give you just one more tip additionally, I highly recommend you to use your hands while studying. What? Yes, I mean a pencil and a note. This could sound ridiculous but yes, it’s important. I saw some people just watching videos without writing down anything when taking lectures. I guess this is because they can reaccess to the content whenever they want. If you’re smart enough to digest all those materials without a note, then hats off to you. But if you’re an average person like me, then note-taking will help you to comprehend the courses much more.

These are the strategies enabled me to finish Andrew Ng’s deep learning specialization in such a short time. I hope this could give you new inspiration for learning method. But undoubtedly, the most crucial factor is dedication. You have to pour yourself into it. I just spared all my weekend for this course. Cause It was so interesting and fun that I couldn’t stop it. Yes, I also think I’m a geek. Yeah 😏😏

So what’s my next plan? Well, putting what I’ve learned into practice!

Thank you for reading and I hope you found this post interesting. I’m always open to talk so feel free to leave comments below and share your thoughts. I also post other DS resources weekly on LinkedIn, so please follow me, contact me and reach me out! I’ll come back with another exciting story soon. Stay tuned!