Week 6

I. ES Prototype

I learned more about Windows programming this week than I did about neural networks. I am now able to run Prof. Rana's code from a Windows program, and to alter the mu, lambda, and plus/comma parameters through a dialog box. Here is an admittedly crude screen shot:

I found some useful sources of benchmarks for neural networks. Carnegie Mellon's nnbench is available here, but the most extensive repository is the one maintained through the Delve project, described here.
I found two relevant papers in the IEEE Visualization Conference proceedings, both having to do with presentation of trees. The first, "Tree Maps: A Space-Filling Approach to the Visualization of Hierarchical Information Structures", describes a Venn diagram-like approach that always uses 100% of the display space. The second, "Cheops: A Compact Explorer for Complex Hierarchies", presents a quite clever approach whereby nodes in the higher levels of a tree are overlapped and accessed selectively, which may alleviate some of the family tree scalability problems that have already been discussed.