bought a Traxxas E-MAXX for $310.55 plus $6.99 shipping

I have seen other RC trucks that have been turned into robots, and the E-MAXX was the fastest strongest best biggest truck that I could find. I need a laptop to put on it. It’s 19.125” long, 14.8” wide, 10.5” high, has 4” ground clearance, and a two speed transmission that lets it go slow and strong or over thirty MPH in high gear.

discovered a way to greatly improve new learning algorithm

This was probably the simplest, easiest solution that I found. The problem was that the outputs of neurons that had a lot of inputs, like 2500 inputs, were always almost zero, but the outputs of neurons that had a few inputs, like ten, would vary a lot. This is because the more random numbers there are (positive and negative), the closer to zero the average will be. I fixed this problem by changing the number eight in this equation to the number of inputs.

n = ((3 ^ (8 * x)) - 1) / ((3 ^ (8 * x)) + 1)

Logitech Quickcam came in the mail

made it possible to save information packets for the neural network program

made the ANN software an Active X EXE so that it could be shared to other programs

After I did this I could have two programs that would have neural network class modules and control them. Both programs would run in parallel. I taught a neural network to recognize whether a ball was to the right or the left, and then loaded it into another program. This program had a label who’s caption would say either “Right” or “Left” depending on what the network’s output was. I could connect the network to the Logitech QuickCam, and then connect motor to the camera and plug it into the parallel port and control it through the one program, and the camera would watch a ball as it rolled by.

bought a Logitech QuickCam on eBay for $7.00 plus $7.00 shipping. also gave the output layer of the neural network the original activation function (threshold level).