Machine Learning Coursera - Andrew Ng
By Jan Van de Poel on Feb 13, 2018
Course: Machine Learning
By: Andrew Ng - Stanford (Coursera, ex-Google Brain, ex-Baidu, deeplearning.ai, landing.ai)
Even if you are fairly new to machine/deep learning, chances are, you have already heared about Machine Learning Coursera from Andrew Ng. Since August 2011 1.8 million students have enrolled in this course, which was also Coursera’s first, and for many, it provides them with the first steps in the field.
I personally started off with no real specific expectations other than to hopefully demystify the magic surrounding machine learning. And although it took me some tries to get to the finish, it was totally worth it.
Why invest your time?
If you are looking to get a good grasp on the basics - that you will likely use on most of your machine learning endeavours - this is a sound place to start. You will spend some time going through the material and exercises, but it will totally be worth your while, here’s why:
Learn from the best
Andrew Ng is an exceptional teacher. Period.
Calmly and gently, he will take your mind on his train of thoughts and explain hard(er) concepts in an simple and intuitive manner. While you are watching the videos, most of the content will seem like it comes natural. For me, only when I had to finish the exercises, it became apparent how well it was explained, yet I had to internalize the content better by revisiting the material presented.
If you need more time to revisit the material, don’t worry.
Demystify machine learning
If you have no general idea how it works, machine learning can seem rather mysterious. The course explains how we define what it means for machines “to learn”, and which types of learning exist.
You will get to understand how computers are starting to achieve these results and an intuition for what machine learning is and isn’t.
Fundamentals to build on
Like most things, the fundamentals are key and a strong foundation allows you to build on top. You will learn about and experiment with the basic building blocks that power the current state of the art techniques.
Although it is highly unlikely you will ever implement these parts in the real world - luckely, you can stand on the shoulders of giants - it is usually a good idea to understand what and why you are doing the things you are doing.
Advice to last a lifetime
There is some interesting material on intuition and evalution of (your) machine learning solutions. Training machine learning in the real world, usually consists of three parts:
- define and implement your model,
- find and prepare the right data for training,
- fire up and run the machines to effectively train your model.
These activities can be time (and money) consuming, so it is good to know how to evaluate whether you are moving in the right direction. You will develop an intuition for how to improve your solution. Do I need more data? Is my model too simple/complex? Is my algorithm/model improving?
With a hint of math
Regardless of how you look at it, math/statistics underpin a lot of what makes machine learning tick. The course will give you an introduction to the basic math required for machine learning, without going into too much detail. So do not worry if you are not a math genius. All necessary concepts will be explained with adequate theory and examples, so you will be able to implement them later on. Depending on where your machine learning path leads, you will need less or more math. The good news is, you can achieve good results, even if you are not an expert mathematician.
Brain on, hands-on.
Each week you will have to apply what you have learned in both a quiz as well as some graded programming exercises. Implementing these machine learning solutions can be challenging, but forces you to think about what you (should) have learned. Some weeks will be easier than others, but when the going gets rough…
To be honest, I feel it’s rather uncertain you will use the code you’ve written here in practice, but I’m convinced it helps you build your skills.
Amazingly, the entire course is available at no cost. You can try as much as you want, and take all the time in the world to finish. Once completed, you can optionally buy a certificate of completion to add to your LinkedIn/resume.
The course itself is a considerable investment of your time and will exercise your brain from time to time, but it will be totally worth the effort if you’re intrigued by machine learning and need/want to learn more. If you get stuck, there are people on the forum to help you out.
You’ve got to start somewhere, and this is a good place to do so.