I was always fascinated by the topic of Machine learning until I decided to teach myself how to do it. So I came through a course provided by Stanford published online. However I was shocked of the amount of math it contained. So what is the mathematical background I should have to be able to understand the algorithms of machine learning? Are there any libraries that abstracts all the maths and focuses on actually designing a software capable of learning?
Much like 99% (don't quote me on the number) of the computer-science related topics, the theoretical foundation of Machine Learning often involves a lot of math... none-the-less, it shouldn't be that difficult to pick up on some basic ML algorithms even without in depth knowledge of calculus.
There are various machine learning libraries out there:
I would say that you should start by trying to build your own simple ML algorithm: maybe a Neural Network or a Genetic Algorithm. Successfully building one will make quite a difference in your understanding... especially since given a specific problem, you might have to customize the ML algorithm quite a bit. Knowing how it works, from the ground up, is going to allow you to make any modifications that you deem necessary.
Programming Collective Intelligence: Building Smart Web 2.0 Applications by Toby Segaran is an AWESOME book! Toby builds simple implementations of all-times Machine Learning classics: Neural networks, support vector machines, genetic algorithms, clustering. All that with simple explanations on how and why they work. As a bonus, all the examples are in Python! But even if you don't know Python you'll understand the book. I highly recommend it
See here for some background texts on machine learning: http://bumphunting.blogspot.com/2009/07/what-are-good-prequisite-textbooks-for.html
Core:
- Linear Algebra (learn to use vector and matrix, factorization, eigenvalue)
- Numerical methods (solutions to matrix equations, and introduction to some simple numerical optimization algorithms)
- introduction to probability (and conditional probability)
Depends on prof.