Machine Learning: road map for beginners

There is no doubt that Machine Learning is one of the hottest subjects today, and there are more and more people who want to learn about it. So I would like to create this post which aims at providing a road map for complete beginners who want to start their career in Machine Learning. If you’re already an expert, just skip this.

First of all, if you surf some big Facebook groups in AI & ML, there are many questions like: « How can I get started in ML without basic math?« , or « I’m not good at coding, can I learn ML? » These are very normal because we all understand that today many ones want to add an ML certification to their CV. But, if you want to be come an expert in ML, or even want to reach an intermediate level, the common answer for such questions is: NO.

In fact, today there are a number of libraries and frameworks such as sci-kit-learn, keras, or TensorFlow which allow you to build a predictive model in just some lines of code, but if you don’t understand the algorithms’ principles, you may get stuck in choosing the right models and the right parameters, which may waste you a lot of time. For example, Normal Equation can help finding the best fitting hyper-plane in Linear Regression but why in many cases we should use some gradient-descent based methods? Or when using gradient descent, what is the impact of the learning rate? Or in Neural Network, is adding more layers always better? Why do we need to randomly initialize weights before training network, and how they should be generated? Etc.
In addition, if you may want to do some AI researches or do a PhD thesis (I did so), understanding algorithms and implement yours are very essential for reading and writing research papers.

BLACK FRIDAY: 92% discount on « AI & Machine Learning from scratch » course by Dr. Long Nguyen (offer valid until 04/12/20)

=============================================
FIRST STEP: Prepare a mathematical background!

If someone says that he wants to start learning ML without math, he does not seriously want to do it!
Indeed, you don’t have to be a mathematician to learn about ML, but some basic background in Linear Algebra, Calculus, and Probability would help you learn ML as fast as a rocket.

So, if you already have a good background in the above math subjects, then congratulation, you’re much more lucky than many other starters. But if you have not, this course on Coursera would not be a bad choice. That Coursera specialization will provide you a sufficient background in Linear Algebra and Calculus.

For learning Probability, if you don’t not have enough time for this very comprehensive course on EdX, then you can have a look at the following compact note from Stanford university.

=======================================
SECOND STEP: Prepare a basic coding skill

There is no slot for non-coder who wants to be a specialist in ML. Otherwise my mom would apply for one!

If you already have a solid coding background, then congratulation again. But if you do not have it yet, just pick a programming language to learn. As the matter of fact, Python is emerging as the most suitable language for learning and for doing ML, but depending on your specific projects, you may want to choose C++, Java, or anything else. If you want to get a quick start in Python, then The Python Bible course is among of the best choices.

=========================================
THIRD STEP: Take a Machine Learning course

This should be an ML course from scratch which teaches you step-by-step the fundamentals of ML, rather than the ones which teach you how to use just three lines of code to build an ML model! Those courses will be for a later step, not now.

This course from Udemy is a very comprehensive course from scratch which teaches you step-by-step to the most fundamentals in Machine Learning. Concretely, you will be walked through the most important algorithms with practical exercises on real-world problems using Python. Moreover, it also covers other AI’s branches such as Fuzzy Logic and Evolutionary Computation, which will be very useful for your study and research on AI.

============================================
FOURTH STEP: Take a more advanced ML course

Once you’ve got the fundamentals of ML, the next step is to find some more advanced courses such as this Deep Learning specialization on Coursera which focuses on Deep Neural Network, Convolutional Neural Network, and Sequence Models.

You can also look for some special ML courses that focus on your specific problem such as Computer Vision, Natural Language Processing, or Data Science …

==================================================
FIFTH STEP: Take an Applied Machine Learning course

Now you can look for some kind of hand-on courses that teach you how to use the most powerful ML libraries and frameworks such as sci-kit-learn or TensorFlow which allow you to build ML models in just some lines of code. Finding an ML book is also a good choice: I would recommend you this Oreilly’s hands-on Machine Learning book.

Trust me, once you have a full background in ML, using such frameworks now makes you feel like a boss! You know everything, you control everything.

BLACK FRIDAY: 92% discount on « AI & Machine Learning from scratch » course by Dr. Long Nguyen (offer valid until 04/12/20)

=====================================
SIXTH STEP: Practice real world projects

It seems not to be easy to get a job without any year of experiences, so you have to build your own. Kaggle is an ideal place to test your ML skills. Join it, try to play with some data sets and challenge yourself by joining into some competitions. Then you can propose some freelancing works in order to build your own experiences. However, it would be always better if you could find an ML job after all.

========================
SEVENTH STEP: Research

To become an expert, you don’t need to be a researcher, but it is one way to achieve that. AI & ML still need a lot of researches, according to the rapid growth of the number of scientific papers in recent years. So if you really want to become an expert in ML, a good way is to read regularly scientific papers to update the latest approaches, and why not try to propose something new by yourself, nothing impossible! If you want to find some papers, Google Scholar is a good place to start. You can also try to look for some PhD offers if timing budget is not your big problem.

=========================
Conclusion

So, above is the 7-step road map for complete beginners who want to start their career in Machine Learning, according to my opinion. You may want to propose your own way, of course.
If you start from zero, it may take you one year from step 1 to step 5, in order to get the basic level. The sixth step should be taken during at least one year then you will get into the intermediate level. Finally, to become an expert in this domain, it will not take you less than three years. That means, from zero to hero (expert), be prepared for the next five years or more, except you are a Mensa’s member!

There is no easy way. It’s normal. But I believe, and I hope all of you are doing and will do so well.
Thanks for reading this so long post, but if you find this useful, feel free to share it.

BLACK FRIDAY: 92% discount on « AI & Machine Learning from scratch » course by Dr. Long Nguyen (offer valid until 04/12/20)

Votre commentaire

Entrez vos coordonnées ci-dessous ou cliquez sur une icône pour vous connecter:

Logo WordPress.com

Vous commentez à l’aide de votre compte WordPress.com. Déconnexion /  Changer )

Photo Google

Vous commentez à l’aide de votre compte Google. Déconnexion /  Changer )

Image Twitter

Vous commentez à l’aide de votre compte Twitter. Déconnexion /  Changer )

Photo Facebook

Vous commentez à l’aide de votre compte Facebook. Déconnexion /  Changer )

Connexion à %s