~ Homework ~

Feed


Homework 1 (Due on Nov 20)

Homework 2 (Due on Dec 25)

K-means for image clustering. Dataset is available on THE MNIST DATABASE. In this project, you need to:

  • Combine both the training set and test set into one set and randomly pick 50000 images;

  • Use K-means and K-means++ to cluster the images you have chosen in Step 1 into 10 different classes. Compute the error rate using the ground truth and compare the performance between K-means and K-means++;

  • Apply Fuzz C-means method in C-means to cluster the images you have chosen in Step 1 into 10 different classes and compute the error rate using the ground truth. Compare the result with that of Step 2.