Entry level posts for machine-learning

Question

Hi,

I am looking for some entry level posts about machine learning. Can anyone suggest anything for somebody new to this subject?

Cheers

Paul

Answer 1

Machine learning is such a broad topic. I would start with Wikipedia and focus in on areas that you find interesting.

Also, you could visit the Stack Exchange site for machine learning.

Answer 2

By 'posts' i'll assume you mean any resource available online.

I recommend two groups of resources:

First, find Machine Learning blogs in which the blogger's preferred language is the same as yours. In my experience, reading a blog post on a single subject (e.g., SVM) while reading through the author's source code supplied along with blog post is about the best way for a programmer to learn ML. A couple of excellent examples are the blogs Smell the Data (Python), and Igvita (Ruby). Both contain (at least) several posts each describing, tutorial-style, specific ML techniques, which include close walk-throughs of their (posted) source code. Igvita, in particular, has excellent tutorials with working Ruby code on Support Vector Machines, Decision Trees, Singular Value Decomp, and Ensemble Methods--like, the other blog i mentioned, an upper-level undergraduate course could be taught based solely on the ML posts in either blog.

Second, I highly recommend VideoLectures.net.

This is by far the best source--whether free or paid--i have found for very-high quality (both w/r/t the video quality and w/r/t the presentation content) video lectures and tutorials on machine learning. The target audience for these video lectures ranges from beginner (some lectures are specifically tagged as "tutorials") to expert; most of them seem to be somewhere in the middle.

All of the lectures and tutorials are taught to highly experienced professionals and academics, and in many instances, the lecturer is the leading authority on the topic he/she is lecturing on. The site is also 100% free.

The one disadvantage is that you cannot download the lectures and store them in e.g., itunes; however, nearly every lectures has a set of slides which you can download (or, conveniently, you can view them online as you watch the presentation).

A few that i've watched and that i can recommend highly:

• Semi-Supervised Learning Approaches

• Introduction to Machine Learning

• Gaussian Process Basics

• Graphical Models

• k-Nearest Neighbor Models

• Introduction to Kernel Methods

Answer 3

Stanford published videos and materials from a set of engineering courses at http://see.stanford.edu

One course by Andrew Ng focuses on Machine Learning techniques http://see.stanford.edu/see/courseinfo.aspx?coll=348ca38a-3a6d-4052-937d-cb017338d7b1

The course is also available on iTunes U

Its a really good course from someone who obviously knows the field well, but he spends alot of the time deriving mathematical results - so if your rusty in linear algebra or prob/stats, you might need a refresher first.