Andrew Ng's course notes are 90% of the way to being a textbook, and are an incredible resource, especially if you're going to watch the lectures.

http://www.stanford.edu/class/cs229/materials.html

I also recommend Andrew Moore's tutorials on Datamining, which is quite related to ML. Found them on HN a few days ago, printed the slides and started reading them. So now I'm just passing it on :)

God, whatever you do please don't start by picking a few datasets at UCI ML site and using R packages to play with the data.

An ideal approach will be to:

- Pick any programming language and start off with plain regression. It may look simple but this will become fantastic base going forward

- Generate a synthetic data set and apply your freshly written regression on it

- Expand your toolkit to include test and training data set generation and calculation of ROC curves and confusion tables

- Add logistic regression, regularizers and other advanced regression models to the toolkit

- Use a real world dataset and develop multiple different models. And pick the best model (choosing the right model itself is a big task in itself)

- Then try coding Neural Networks, SVM, etc.

Agree with you Paras!

Shubhendu here by the way! :)

I was taken aback a little by the suggestion to use the UCI repository with R for beginning ML.

I would agree with your approach, I learnt all my basics from Andrew Ng's course and his course more or less follows what you said. :)

And I prefer the book by Bishop, It strikes a fine balance between the maths and the ideas.

I would add:

Programming Collective Intelligence (O'reilly) http://oreilly.com/catalog/9780596529321

http://ianma.wordpress.com/2009/07/19/machine-learning-for-b...

Practical Artificial Intelligence Programming in Java http://www.markwatson.com/opencontent/

"Programming Collective Intelligence" is very short on theory. It's good for getting your feet wet and for getting over your fear of something new and difficult, but otherwise without proper comprehension of the phenomenons involved in those algorithms, you're not going to get very far.

Yes, it is short and basic in theory, but I it is very practical. In my case I learn best trying and applying concepts in practice. That is why I found the book very useful. Also I accompanied my learning with some other books (i.e. Machine Learning from Tom Mitchell) and academic papers that filled the theory gap. As a started point to teach the basics of ML and to encourage to go and learn more, I think PCI is very good.

Exactly. Please refer this thread http://news.ycombinator.com/item?id=208892

Nice to see a link to online courses. I've been studying Bishop's book in my spare time.

Undone. I do that all the time. Probably I downvote accidentally more than intentionally. Whatever the reasoning for now allowing us to change our votes is definitely off in this case...

I'm curious what you guys use ML for

MATLAB

I was wondering more about like regression of customer data, creating recommendation engines, or something along those lines.