Introduction
If you're looking for a solid foundation in machine learning, the Andrew Ng course at Stanford University is a great place to start. Recently, I completed the course and I want to share my thoughts on the experience.
My First Thoughts
Firstly, I really enjoyed the course and I believe it provides a great base for any machine learning practitioner. The course has a low-level focus, delving into the math and mechanics behind the concepts. This may be challenging if you don't have a background in university-level calculus and linear algebra, but the course doesn't rely heavily on math when grading. I would still recommend looking into linear algebra to better understand the formulas.
The course covers a wide range of topics, including supervised and unsupervised learning, clustering techniques, and good practices for achieving great results. It's also great that the course doesn't focus too much on deep learning, leaving room for that topic to be covered in another course, such as deeplearning.ai.
Programming assignments
The labs are a big part of the course, and I must admit, I had my fair share of frustrations while working on them. I found that Octave, the programming language used in the labs, was different from the high-level and OOP languages I was used to. However, I found that thinking in terms of math and linear algebra, rather than programming, helped me get through the assignments. One tip I can offer is to start using vectors as soon as possible. Not only will this make your code faster by avoiding for loops, but you'll be able to translate math formulas almost directly into code.
Another tip is to take advantage of the tutorial section of the labs. I didn't realize it existed until the last weeks of the course and it would have saved me a lot of time and frustration. I also found it helpful to draw out the concepts, like in my drawings of backpropagation and gradient descent (shown below).
Final thoughts
In conclusion, I highly recommend the Andrew Ng course to anyone looking to gain a comprehensive understanding of machine learning and data science. The course is 11 weeks long, but you can complete it in 3-4 weeks if you're able to devote enough time.
You'll come away with a strong understanding of topics beyond deep learning and good practices for data science. Plus, it's always nice to feel like an expert when Professor Ng says "those of you who are experts in linear algebra will recognize this."
If you're struggling with any concepts in the course, feel free to reach out to me on twitter @ramgendeploy for help. Just keep in mind that I can't help with programming assignments, but there's always the tutorials section in the forums!
Top comments (0)