k_{1}
and k_{2}
so that they are neither too large nor too small when compared to the range of f. Run your ACA for many iterations and watch the ant colony getting groomed. How many clusters do you see? Tune the above parameters until you are satisfied with the results.
For a particular parameter choice, rerun your program a few times starting from different initial conditions. Report your parameter choices and the corresponding results. You can use our sample code that draws a rectangular grid-world, and randomly places a few dead ants and live ants on it.k_{1}
, k_{2}
, and α
that give you meaningful clusterings of the 16 animals. One possible clustering you may expect to see is all the 'terrestrial' data items in one cluster and the 'bird' items in another.α
defines the scale of dissimilarity: it determines when two items should be or should not be located next to each other (p.229). Once you are satisfied with the results from the above, rerun your algorithm by varying just α
. How do higher and lower values of α
affect the clustering performance?Last Modified: April 25, 2024