Ask HN: Best online courses for machine learning? | Hacker News Reader

Ask HN: Best online courses for machine learning?

65 pointsaliirz21 comments

Ask HN: What are some of the best online resources to jump into practical machine learning

Contrary to popular opinion, I find that Andrew Ng's Intro to ML (on Coursera/Stanford) focuses too much on basic math and theory - which doesn't detract from the course's quality, but makes the course a drudgery to go through.

Programming exercises involve a single line or two, and that too in Octave - which was all the rage back when the course was launched, but it's not so useful now.

Instead, start with this - https://www.fast.ai

It emphasizes practicality to the extreme - you are only taught theory/domain knowledge when needed. The instructor's amazing, the massive scale of knowledge imparted boggles the mind, and you feel like you've accomplished something when you're finished with it.

Best of all, it's free. And you can start Deep Learning there too when you're done with ML, if you feel the need (or interest).

I'd second this class. I have done both and if you're brand new to machine learning I'd do fast.ai first. You gain a greater understanding from Andrew's course, but if you're not going to become a practitioner fast.ai is better.

i have to agree with you here. It's good class if you want to focus on theory and veer towards academic part of ML.

There should also be some sort of polls for these sort of questions or at least for mapping goals. I past polls if any in ...yc.com/news feed. How do I go about it?

I really love the code first approach of Fast.ai, I'm impressed by teaching style of Jeremy its clear, succinct, and to the point.

To learn a solid theoretical foundation: Caltech's "Learning from data" [1]. It is one of those rare courses where the professor is so good that he manages to make tricky concepts seem almost trivial (like Paar's "Understanding cryptography").

Look elsewhere if all you want is to learn tools and start practical projects ASAP without really understanding what you are doing. Tools come and go and will serve you for a while, concepts are timeless and will serve you for life. You need both, of course, but I wouldn't skip the theory, specially when such an amazing course is on tap.

[1] https://work.caltech.edu/telecourse.html

Agreed! 100%. What level of math is required for the LFD course?

Don't worry about the math, one of the great things about this course is that the professor is a magician when it comes to explaining and defanging scary looking math, you'll be fine. Having said that, it's mostly basic matrix and vector algebra, basic probability and basic calculus (all high school level).

Thanks for the words of encouragement! With that said I will watch the lectures, read the book and do the assignments.

I would recommend this course: https://mlcourse.ai/. It starts on February 11.

PROS

- It has a good balance between theory and practice.

- There are lectures covering the theory and practice.

- There are practical assignments you need to code with Python.

- It includes in-class Kaggle competitions.

- It includes a rating system so you can compare your progress with other students.

CONS

- It has some prerequisites. You need to know Python(at a basic level) and some basic knowledge of math(calculus, linear algebra, etc).

- It is a difficult course. You will need between 5 to 10 hours each week for assignments. Each week is usually harder than the previous one.

You can find more details in this post: https://www.kaggle.com/general/77771

Google ML crash course: https://developers.google.com/machine-learning/crash-course/

Facebook field guide to machine learning: https://research.fb.com/the-facebook-field-guide-to-machine-...

Training on Machine Learning with AWS: https://aws.amazon.com/training/learning-paths/machine-learn...

Can anyone comment on whether these are worth the time investment versus more traditional courses?

I haven't taken any of these courses. There were these courses from Amazon https://aws.amazon.com/training/learning-paths/machine-learn... that were highly recommended. But it all comes down to what you really want. If you just want to get a bird's eye view of the ML algorithms buffet for you to implement one of their Black box models then they are a good place to start. But for anything deeper, I would go with Charlysl's recommendation.

Udemy's Machine Learning A-Z course: https://www.udemy.com/machinelearning/

All lessons in R and Python. Wide range of content.

This'd be a safe bet https://www.coursera.org/learn/machine-learning

The online Stanford courses--CS224, CS229, CS231--are an excellent introduction into modern AI. CS231 with Andrej Karpathy in particular was a game-changer for me. It has three very thorough and well-designed assignments that will have you implement many of the basic algorithms discussed in the course.

The pre-requisite linear algebra for these subjects can be learned through Gilbert Strang's MIT course. A basic grounding in statistics and probability theory, along with calculus will also help.

This one: https://see.stanford.edu/course/cs229

It's a recording of CS229 at Stanford. This course is much harder and more thorough than the one on coursera.

There's a free intro course here: https://www.udemy.com/machine-learning-intro-for-python-deve...

This depends upon you first telling us at what level you are? beg, int, ad.

Not OP but I'm a beginner. Any recommendations different than what was already said?

fast.ai is great.

Reply on news.ycombinator.com