I got the second Logitech QuickCam in the mail. Now I’m trying to hook it up to the neural network program.

Sold the E-MAXX for $242.50.

Today I won a second Logitech QuickCam Express on eBay. It cost $9.99 plus $7.09 shipping.

I perfected the motor control circuit.

I designed and built a motor control circuit.

I asked a potential sponsor if they would like to sponsor me. VIA, a company that sells very small computers, is looking for teams that want to use Mini-ITX mainboards for the main control of their robots. These appear to be designed mainly for multimedia presentations…

I changed the numbers in the transfer function lookup table to doubles for higher precision.

I made it so that the transfer function is drawn on the user interface, and all of the neurons’ outputs are shown. After that, I made it so that the different variables that control the steepness, range, and resolution of the transfer function can be changed while the program is running. Now it is much easier to see the results of my adjustments.

Also, to solve the problem where the transfer function would try to look up a number whose index didn’t exist, I changed the range of the lookup table so that all of the neurons’ outputs would be somewhere on it. I then discovered another problem: all of the numbers at the end of the lookup table were 1, and all of the ones at the beginning were –1. The point in using a sigmoid transfer function instead of a threshold function was to make it so that the output always changed if the input was changed. The 1/-1 problem makes it so that it won’t always change.

Today I finished the final draft of my research paper.