What are some tips and tricks for getting through the Deep Learning Specialization?
As someone with a non-quant background, the key for me has been finding concepts I’m passionate about. I’ve been excited about the possibilities Deep Learning predictive/classification models since I learned about them. That excitement changes my mindset. When I hit a roadblock, it becomes an opportunity to learn more or to connect with someone who can answer my question. When I’m not excited, a roadblock is confirmation I should learn something else. So, my first piece of advice is find something that gets you excited. You can read blog posts, explore these forums, watch YouTube videos, or...ask people!
If you are a practical learner, like me, theory can be tough. I learn backwards, I need to know how something is applied and why/when I would use it to fully grasp learning concepts. So, I found that experimenting helped me sustain my learning. I did small projects using Kaggle datasets and I maximized Microsoft Azure’s free ML Studio credits. Try to find ways to make what you are learning applicable to problems you’ve tried to solve or think you might solve. That helped me.
For me, the best way to tackle the specialization was to develop a project alongside the courses. Think of something that interests you - in my case it was painter classification, but it could really be anything, as long as it interests you. Then turn your interest into a project, set it up and apply the content of the videos and the concepts you have learned to your own project. It's exciting and it forces you to understand what you have learnt in depth, as you will most likely have to alter it slightly. If you haven't got the time to do so during the course (for example because you are working as well), do the course and afterwards apply the concepts to your project. You will then most likely have to go back at times and review videos etc., which will also help you to understand the content more. I also find that doing so will keep you motivated, especially as your own project will improve and achieve better results. And as it improves, you will develop new ideas in order to try and make it even better. Which will again force you to come back to the specialization and/or even other sources, solidify what you have learnt and/or learn and apply new concepts.
Never get stuck for more than 10 minutes if you run into challenges. If you cannot figure an answer out within 10 minutes, either in the programming or quiz section, get back and review the content. This will help your learning a lot. If this doesn’t work, consult the forums. I guarantee you that you will find an answer there. If it’s not there, create a new forum entry, and the mentors will get back to you shortly. Too often I’ve witnessed learners being stuck at a specific point for hours. These hours don’t support your learning progress, they just frustrate you. Avoid it through reviewing the lectures and browsing the forums.
Also, try to make a project plan before starting the course. You’ll know exactly how long the video sessions are and how much time you should plan per exercise. Always try to watch the lectures and complete the exercises in one sitting.
I’ve compiled a few learning resources, hoping they will help you.
How to finish Andrew Ng’s first Deep Learning course in 7 days: https://medium.com/machine-learning-world/netflix-or-coursera-how-to-finish-andrew-ngs-1st-deep-learning-course-in-7-days-6fa293ee83d8
The Deep Learning(.ai) Dictionary: https://towardsdatascience.com/the-deep-learning-ai-dictionary-ade421df39e4
Convolutional Neural Networks for all: https://towardsdatascience.com/convolutional-neural-networks-for-all-part-i-cdd282ee7947
I would say: Commit, Keep Moving, Loosen Up Your Mind and Remember your Purpose for Learning
There will be days when you don’t feel like watching yet another video, or finishing another assignment...but try to complete something, no matter how tiny, each day, once you break the chain of habit it gets more difficult to get back on track.
There will be days when you feel you’re not getting 60%, let alone 100% of what is being said. Don’t get stuck and try to move on, sometimes some of my doubts in a current chapter where answered in the next one, if I had gotten stuck I would probably still be there.
Try to relax your mind, some of the concepts are highly abstract and not something you can apprehend as you would learning about how to hammer a nail, just try to go with it, use your imagination and some analogies to help bring those concepts home.
Whenever your motivation is flinching just remember what led you to start this learning path. For me it was those endless public debates and coffee shop conversations, in which I felt everybody was talking about a topic that nobody really understood. And worse, saying it would have a major impact in everyone’s lives. That got me pumped!
Deeplearning.ai’s online courses have been phenomenal for me, I loved completing them and learnt a lot. Now that I help out mentoring them I see a few of the challenges with the Coursera technical set up.
One trick I learnt and which can save you heartache is to maintain a txt document while you are completing the exercise iPython notebooks, copy each of your code snippets into this document as you write them. This has two advantages
1) you have a record of your code/the exercise more easily available for later reference and
2) that one time you come face to face with the iPython Kernel crashing and have to restart it to be faced with a fresh empty notebook you will have a record of your current worked solution to use to get back up and into the flow.
when I started the first course in the specialization, I felt it's my journey to use mathematics in the correct path, it's my dream, thanks to AI!
I understood the contents deeply, especially that I'm a mathematician, I like to brainstorm ideas and implement the theories while I'm learning, I wrote general notes on a copybook, and start creating my own projects based on what I have learned.
I spent about 2 months the 5 courses in the specialization, the most important thing I noticed was to implement everything I have learned, and to improve my skills by exploring projects and start building neural networks for impact-full projects.
Deep Learning is awesome! I'm proud to be a student of Sir Andrew Ng and a member at deeplearning.ai.
The simple tip for getting good in deep learning is learn deep learning from the good training center from where you get the opportunity to work on live project. To become a Specialist in Deep Learning you have become in making algorithm faster as well as good in mathematics.
The first pre-requitics is learning deep learning technology from good training center.