CS 531/ECE 517 Computer HW 4

Due (Friday) May 11

Consider the MNIST Database of handwritten digits. There are a LARGE number of samples in both the training and test datasets. You might consider writing your classifier using a small subset of each, until you are sure it works. If you do use a subset for the assignment, be sure to report what you did. You need to deliver both the code and a write-up of your results.

1) Implement a Support Vector Machine using a Quadratic Programming Solver.
a) What is the error rate?
b) What was the run-time for this classifier?

2) Compare the results using various kernels
a) Gaussian. Identify the ideal radius (gamma).
b) Polynomial.
c) Radial Basis Function.
d) What are the error rates for each?

3) Extra Credit 1. Identify or create a test dataset of your own. Choose a 2D subset of the data.
a) Visualize (make an image of) the decision boundaries and margin samples.
b) Show the difference in decision boundaries for various kernels

4) Extra Credit 2. Identify a new kernel that was not discussed in class (gaussian,polynomial,RBF)
a) What is the kernel? Describe its "profile", how does it work.
b) What Phi-space does it compute the inner product in?
c) How does it compare to the other kernels?
c) What is its error rate for each dataset?