Estimated read time: 3 min.

Just finished Andrew Ng’s Machine Learning course on Coursera, and it’s GREAT! Here some thoughts and observations:

What’s great about it

  1. Well designed learning curve

EVE Online game’s (in)famous crazy learning curveEVE Online game’s (in)famous crazy learning curve

This is especially important for people that never heard of Machine Learning. Not assuming the student have any prior knowledge and gradually guide them through complex concept makes the learning experience challenging yet still fun.

  1. Avoid complex math, yet find a way to still enable student to do ML (meme)

The Andrew Ng ‘Silent Protector’ MemeThe Andrew Ng ‘Silent Protector’ Meme

Maybe the biggest fear for people want to get into Machine Learning and AI is ‘I’m not a math person’. Being able to not getting into too much math yet still explain clearly the concept is invaluable, especially for totally green guys.

  1. *Octave/Matlab is more ‘math’ like, less digression on programming language itself*

Some people might not agree with me on this. Yes Octave/Matlab doesn’t have all the fancy libraries like scikit-learn and Pandas, yet it’s very expressive when it comes to represent math equations. Transfer equations from class to Matlab code is easier than Python IMHO.

  1. Cover most popular models, good foundation

  2. Linear/Logistic Regression

  3. SVM

  4. Neural Network

  5. Collaborative Filtering

  6. Anomaly Detection

  7. K-Means

  8. PCA

With all these algorithms/models under your belt, you are ready to solve a lot of problems with Machine Learning.

  1. Provide practical ML projects knowledge, not only algorithm and programming Besides theory, the course also offers very practical Machine Learning project knowledge, hot to build a pipeline, how to structure the problem solving, etc.

  2. Well designed quizzes and assignments, as part of learning too

I was always amazed by how well the quizzes and assignments are designed. They are challenging, yet with some efforts achievable, and at the same time offer some new perspective on the lesson. I always learned a few new things doing those and totally enjoyed them.

What’s lacking?

  1. A bit dated on libraries and architectures

You won’t find the high-level Keras, TensorFlow or PyTorth here, but this course is about foundation of Machine Learning and it delivered on its promise. 2. Could use more examples/applications of ML for motivation

There are a lot of exciting development and applications on ML/AI field. If students could be exposed to more of those, it will give them more reasons to keep learning.

Final Thoughts

‘Don’t worry about it if you don’t understand’ ™️‘Don’t worry about it if you don’t understand’ ™️

Overall great course if you are totally new to Machine Learning. All of the well thought out contents coupled with Andrew Ng’s gentle and calm explanation makes the learning experience a breeze and a pleasant journey. The road ahead for Machine Learning might not be so smooth after all but having a ‘soothing’ start could carry you a long way. 👍


Michael Li Avatar Michael Li is the creator and lead developer of this site.

Published

Category

Machine Learning

Tags

Stay in Touch

Get Monthly Updates