NOTE:  The primary objectives of this listing are to provide the teams with feed-back and a sense of their relative performance.  The teams were evaluated on the basis of the progress made towards a correct implementation.

1st:  University of Alabama-Birmingham

Good	Bad
- logic completely correct	- TSS computed incorrectly and too often
- training by epoch	- can not handle variable number of layers
- better understanding of bias	- less elegant use of MPI primatives
- less hardcoding, more robust

2nd:  Georgia Institute of Technology

Good	Bad
- great documentation	- training by pattern
- good use of BCAST	- infinite while-loop with no break
- good flow control and structure

3rd:  University of Kentucky

Good	Bad
- great use of BCAST / Reduction sums	- training by pattern
- very elegant solution	- incorrect use of unit biases
	- overwriting weight array with delta wij's
	- inner loop conceptually incorrect

4th:  University of South Carolina

Good	Bad
- good use of structs	- no dynamic memory allocation, unused statically declared memory
- attempted modular code	- weights randomly initialized in slaves, not master
	- main iterative code not in main iterative event loop

5th:  Christian Brothers University

Good	Bad
- great dynamic memory allocation	- each iteration sending of params, not needed
	- weights sent too frequently and incorrectly
	- training by pattern
	- never applies the delta wij's
	- no implementation of feed-forward or back-propagation code

6th:  University of West Florida

Good	Bad
- good explicit handshaking	- no implementation of feed-forward or back-propagation code
	- very little code implemented
	- training by pattern

7th:  Western Kentucky University

Good	Bad
- interesting initial weight distribution	- no use of MPI for communication
	- very little progress made

8th:  North Carolina A&T

Good	Bad
- used competition as a learning experience	- very little progress made

9th:  University of North Florida

Good	Bad
- attended the session on Friday night	- did not return on Saturday