Hello Raj,

in short:

Classification algorithms use linear or non-linear decision functions for the separation of different classes. For multi-dimensional problems where these methods are not applicable, SVMs create a sub-space representation of the features to allow a the application of classification algorithms. Thus they are helpful for the classification of complex and multidimensional machine learning scenarios.

best

Benedikt

If you want a very basic introduction, you can start by looking at the material of some courses. Some examples are the lectures by Andrew NG:

http://see.stanford.edu/materials/aimlcs229/cs229-notes3.pdf

Or the "Statistical Learning Theory and Applications" course at MIT (Lecture 6):

http://www.mit.edu/~9.520/

[I can't resist to cite my own small seminar on kernel methods, particularly lectures 3 and 6:

http://ispac.ing.uniroma1.it/scardapane/teaching/neural-networks-a-a-2013-2014/]

Starting from here, you can either practice with some existing implementations, or start deepening the theoretical aspects. For the former point, there are a wide range of toolboxes and libraries:

http://www.kernel-machines.org/software

Before experimenting, I would also read the following nice guide on the practical aspects of classification using SVMs:

http://www.csie.ntu.edu.tw/~cjlin/papers/guide/guide.pdf

For the latter point, here you find a selection of important papers:

http://www.kernel-machines.org/tutorials

I would advise to start from " A Tutorial on Support Vector Machines for Pattern Recognition" by Burges and then move on to more difficult ones.

An alternative is to refer to a more comprehensive book, such as "Learning from Data" by Chekassky, the classical books by Vapnik, the introduction to Machine Learning by Alpaydin etc.