Machine learning & mathematics - I am not from a statistics (or mathematics) major background. Just bachelors in computer science but had learnt algebra, probability, statistics, Integration & Differentiation, Permutation & Combination, etc. 
Over last ~3 months, I've made my hands slightly dirty with ML code and concepts. I've started doing few sample exercises and reading blogs/wikis and articles.
But, is it really important to learn the mathematical notations to get a solid grasp on the internal concepts? 
For ex. SVM Wiki . I have some idea about Kernel Trick but wanted to learnt bit of internals, but it completely blew me off.
How important it is to learn the mathematics behind the concepts?
Please share your thoughts on this.
RB.
 A: 
How important it is to learn the mathematics behind the concepts?

It depends on what you value.  A lot of data scientists I have worked with have not taken the time to learn how an SVM solves the optimization problem it solves.  Nor could they explain the details of maximum likelihood estimation in a more rigorous way than "find the maxima of the likelihood function".  However, that did not stop them from being good at their job.
Not learning the mathematics behind these techniques is not a death sentence (I personally couldn't tell you exactly how SVM does what it does), however it might limit your ability to solve more nuanced problems in the future.  People who tend to learn how to use ML without understanding how it does what it does often go on to think of ML as a unilateral solution to any quantitative question. My advice is that you need not know EVERYTHING before using ML, but you should be capable of learning what how an algorithm does what it does if you were required to know so.
