csc 591-024, (8290)
csc 791-024, (8291)
fall 2024, special topics in computer science
Tim Menzies, timm@ieee.org, com sci, nc state

Install Python 31.3

• for linux
• for mac
• for windows: when you work it out, tell the class

Get the extension extension going

• remember to use the 24Aug14 branch
• see examples here
• note that data divides into small,medium, high dimensionality (number of x columns) and size (number of rows)

Here is a command line that runs all the current built-in examples. Please run it

python3 -B ezr.py -D -e all -t data/optimize/misc/auto93.csv > ~/tmp/out

Run this code and answer the following questions. Write short answers for each. Submit one set of answers per team.

1. heavens d2h is short for “distance to heaven”. How is it calculated?
2. chebys : how is the cheyshev distance different to d2h?
3. likings
   • in english, explain how loglike is calculated? and how is that calculation different for numeric and symbolic columns?
   • Diversity sampling means that the next thing we look out should be different to everything seen before. So explain: “selecting for min loglike is a synonym for diversity sampling”
4. mean-vs-median
   • This code recursively divides data by (a) slitting data according to everyone’s distance to two far points; then (b) recursing into each half.
   • What is the difference between half_median and half_mean?
   • Referring to this output from
     python3 ezr.py -e mean_vs_median -t data/optimize/misc/auto93.csv
     does mean or median splits make a difference?
5. python3 ezr.py -e clusters -t data/optimize/misc/auto93.csv generates a tree generated via mean splits
   • reproduce the same output
   • How would this tree be different if we used median splits?
6. python3 ezr.py -e clusters2 -t data/optimize/misc/auto93.csv shows the results of prediction by (a) cluster the data (see clusters) then for each test (b) find its nearest leaf cluster; then (c) using either the median value of that leaf or the 1,2,3,5 nearest neighbor.
   • generate that output and show it in the homeworks response. t
   • Based on these results, what approach would you recommend?