First, if your group has an algorithm that you'd like to code up or a project that interests you, then do that. 

Here are some suggestions you're welcome to try or that may trigger your thinking about something that interests you. 

Perform timing studies on any of the algorithms that we've discussed

1.  k-means clustering

2.  Canopy clustering

3.  EM algorithm

4.  glmnet - regression

5.  glmnet - classification

Here are some algorithms that we've come across but haven't coded up for class

1.  glmnet - multiclass

2.  combined canopy clustering with k-means

3.  combined canopy clustering with EM

4.  Planet (binary decision tree for MR) PLANET-vldb2009.pdf. 

5.  Stratified sampler

6.  Canopy / random sampler initializer for k-Means