Conclusion

• This lecture has used two dimensions for support vectors. This all works in higher dimensions.
• Take Home Points
  1. Learning algorithms can be parametric or non-parametric.
  2. Support Vector Machines make use of support vectors to generate maximum margin separators.
  3. If you can't find good linear separators, you can use the kernel trick to project to higher dimensions where you can find a linear separator.
  4. SVMs are a framework using linear separators and kernels.
• Reading: Read Russell and Norvig's Learning from Examples Chapter section 9 (pp. 755-758) for this week.
• Reading: for next week read the Deep Learning Wiki