The Hunter-Gatherer Java Demonstration

This is a demonstration of the Hunter-Gatherer optimization system used in the Mikrokosmos Semantic Analyzer.. The demo here presents several Graph Coloring problems.

Things to remember:

Optimization problems such as these require a lot of memory. Stick to the smaller examples if you are lacking in that area. Also remember that this uncompiled Java (at least my Java) is about 100 times slower than the compiled LISP code I usually run Hunter-Gatherer on, and probably hundereds of times slower than compiled C code (I'm also working on a C version). Example 5 runs in 3 seconds in LISP and 5 minutes in Java. So be patient!!! One last thing: the larger graph coloring problems shown here are probably more complex than real world problems, which typically can be decomposed to a greater degree. Hunter-Gatherer was designed to find and exploit compositionality. Natural Language Semantic Analysis , for example, can be performed in near linear time.

So, here's the demo. I'm working on making it more interesting (i.e. showing the input graph and tracking the progress, as well as a side-by-side comparison with other optimization methods). The problem being solved is: color the input graphs using as many reds as possible (with no two adjacent nodes having the same color).

Last revised December, 1996. Send questions and comments to Stephen Beale or visit my website.