HN Explorer

Ask HN: How to learn machine learning on my own?

55 points by tixocloud ↗ HN
Hi,

I'm looking to improve my data science skills and have taken Coursera courses.

Are there any additional resources for me to learn on my own? What's the best approach?

My background is in software engineering, web development and business intelligence.

16 comments

[ 3.1 ms ] story [ 45.2 ms ] thread
selmat
There were plenty of similar threads with very valuable resources. Let me mention some of them:

> https://news.ycombinator.com/item?id=11663952

> https://news.ycombinator.com/item?id=11509921

> https://news.ycombinator.com/item?id=11887147

> https://news.ycombinator.com/item?id=198601

> https://news.ycombinator.com/item?id=11883490

Best google phrase is: "machine learning site:news.ycombinator.com"

I have list of resources I am going to read but day has only 24 hours.

arcanus
The MOOC on coursera is an excellent place to pick up some basic terminology and simple algorithms. It is taught at an undergraduate level, so it is rather accessible.
saiko-chriskun
he's already taken coursera courses and looking for further resources.
ddxv
My style is that I like constantly trying to put myself further into something I don't understand. If you read about something interesting in this forum, just go straight into trying it out. Install everything you can, test out the languages, packages and tools to the best of your ability.

The more you don't understand, the better you're doing! Keep asking questions. Keep trying tutorials. Keep pushing into what you find confusing and difficult.

I can't say for sure this is a great method of learning, but it's the style I enjoy most. Good luck, and thanks for opening this thread.

tixocloud
Thanks for your thoughts. I was thinking about picking a project to work on and using that as my motivation. I find myself sometimes just losing focus after watching the videos. Do you ever experience that?
arsenide
Not sure about OP, but I certainly feel this way. When I want to learn a new language/framework I work on a project using it; instead of watching videos I mostly use the docs and Google/SO when I get stuck or have a code design question, possibly watching/reading any seemingly interesting videos/articles I happen to come across along the way.
ddxv
Yes, I know that feeling too. =)

I get bored too. As soon as I know enough to start my own small project I do it. Even if it's not something I could ever put on the Internet, it is enough that it makes me interested again in understanding what I am learning.

sl8r
What are you looking for in particular?

If you're looking to gain functional familiarity / put in practice reps with classical classification & regression algorithms, I think that running through the online tutorials for scikit-learn is the best bet.

If you're looking for the theory behind the above, I think the book by Peter Flach is the best intro; "Elements of Statistical Learning" is the classic tome, but much more mathematically motivated.

If you're looking for more specialized subjects, each has its own resources. Bayesian modeling? Gelman's BDA3 and Cam David Pilson's github book. Gaussian processes? Rasmussen. Etc., etc. for neural networks, reinforcement learning, etc.

As a random recommendation: David Mumford's "Information theory" is eclectic and fun, but disconnected from the mainstream.

tixocloud
I'm looking to gain functional familiarity but also at the same time working on real world applications.

Did you mean David Mumford's "Pattern Theory"? It was the only book I could find that's close to the topic. Unless I'm looking at the wrong David Mumford?

DrNuke
The best way is to focus on one industry or application (e-commerce, natural language, image recognition and so on) you are really passionate about and start practicing theory and models from there. Machine learning is a bag of tools, the most important point is still the business model asking for it, why and how: target, metrics, execution, validation, deployment.
smcguirk
I just started the book "The Master Algorithm" by Pedro Domingos. The start seemed a bit to much of a sell, but the basic approaches to understanding the direction of this discipline seem sound.
venkasub
Start attempting Data Science competitions; the learning rate just exponentially explodes.

Also, look at the solutions from past exercises, it really helps in how different people solve the same problem with different approaches.