The idea behind Support Vector Machines is simple. Plot your data on a chart and then draw straight lines that separate that data.
Hit '>Play' to learn more about Vector Machines.
In practice this can be difficult. Straight away you run into three big problems.
If data can be separated by a straight line then there are always an infinite number of straight lines that can separate your data. Which one do you choose?
For simple problems defined by two of three variables you can draw a 2D or 3D scatterplot and plot a line or plane to separate the data. But most real world problems have dozens of parameters. How do you effectively separate this data?
Very often the data can't be separated by a straight line or flat plane. How do you separate the data with a straight line or flat plane if the only separation is with a squiggly line or a complex concave and convex hyperplane?
**** Spoiler alert ****
Support Vector Machines - or 'SVMs' - address all these problems.
They will determine the optimum line between data based on maximising the width-of-the-street that separates data into classes.
For complex problems with dozens (or hundreds) or parameters they will determine a hyperplane that separates the data.
If data can't be separated by a straight line or a flat hyperplane (if this is the case data is described as non-linearly separable) then SVMs project data into a higher dimensional space where they can be separated.
The video above describes why we use support vector machines, the concepts behind them, how they work and how SVMs solve the three problems highlighted above.
