Description

Assignment 2

Topic clustering on 20 newsgroup dataset.
Gamma = 0.01 was chosen based on silhouette score
For kernelized k-means, the values are tabulated below

Homogeneity Completeness V-measure Adjusted
Rand
Index AMI NMI FMI
Naive TF-IDF 0.302 0.387 0.340 0.009 0.299 0.342 0.194
Sub-
Linear 0.380 0.464 0.416 0.174 0.376 0.420 0.255
Max TF- normaliz ation 0.192 0.312 0.238 0.045
0.188 0.245 0.175
Bi-gram 0.01 0.286 0.001 0.01 0.001 0.009 0.224