programmed in Snail IDE, Press the dev button to see maths In dev mode, select epochs (default 10000). epochs is the training length, 10000 is about 30 seconds. Press space to continue training after it's already done. Press the dev button again to hide all the maths. If you are not in dev mode: The neural network displays as follows: red lines represent negative values, and white lines represent positive values. The brightness of the lines is how strong the signal is. The lower the brightness, the smaller it is. But it's only bright based on its category height. If it's -1, that line will be red and very bright. If it's just 1, it will be bright white. Basically, the lines display brightness based on how far their number is from zero.
all me. My longest run was 1000000 epochs I got an error of 0 (my computer couldn't count any lower) And at that time, the output could reach the random target within seconds!!! My first project that runs entirely on math! It's recommended to put inputs and outputs to 4 The hidden layers use leaky ReLU the output uses sigmoid. This is my highest score, got a 0.00457 error, which is really good! That means it was a really good run, because that means it messed up almost 0 times. So w1i, w2h, and w3h are the weights. Input is the input, output is the final number, and target is the goal for the output to reach.
