Click the green flag and watch, not interactive except for the RPI. AI weights and biases are random each time. The two dots, F1 and F2, will move around randomly. Whichever one gets closest to T1, the target, will have its hidden neural layers copied to both and modified slightly, refining it and getting more accurate. It doesn't work all the time though. (For moving target versions, not this one) Works best when the target isn't too close to the dots. If both dots start by going in circles, and that doesn't change by the second or third iteration, restart the project. Turbo mode (Shift+Flag) to speed up training B to clear trails
Not just the "point in direction" and move there type of AI. It's got layers and math, because I wanted to do it the hard way to see if I could The lines on the top and bottom are a visualization of the layer Weights and Biases (mirrored, top is F1 and bottom is F2). The three lines on the right are the scale, and the 4 lines are the outputs of F1 and F2. Whichever one "won" is whichever AI got closer and is being duplicated for the next generation. Update 5/31/2023: Added a custom XY grid to improve results. Added tracking lines so you can see the progress (resets every 5 iterations). Added Rounds Per Iteration (RPI) slider, which allows you to change how long the AI runs each time. Recommended range: 20-60
