See Inside You can use this for the brain of any AI based project If you use the idea, or code, then give credit! I have been looking for something like this and no one has made anything like this, so I decided to make it
ALL me @batcom If you're not into math or AI this will be pretty boring Explanation: ok, so first there is the setup block, the amount of numbers with spaces between determines the number of inputs, for example if you put (1 1 1 1 1) in the first input place in the setup block then it would make and expect 5 inputs when training, the second input called "hidden layers" of the setup block determines the configuration of the hidden layers of the network each number with a space in between is a separate hidden layer the number tells it how many nodes are in that specific hidden layer. for example (5 3 2) would make three hidden layers with 5 nodes in the first 3 in the second and 2 in the third. then there is the output spot, where the number of neurons in the output layer is determined in the same way as the inputs are: the amount of numbers there are with spaces in between changes the amount of the neurons in the output for example (1 1 1 1 1) meakes 5 neurons in the output, I don't know if this part works 100% and it may function incorrectly with more than one output, though it does work pretty well with any amount of inputs. then there is the bias, which you set to one if you want bias's to be included or zero if not, I don't know if I implemented bias correctly. and you can change the learning rate which you should know what that is, it's the amount that it alters the weights by each time, closer to zero = more accurate but slower. next we have training, where in the first input you specify what the input values will be for each node in the input layer, where you have each input value separated again, by a space. then the second input in the training block you specify what the output you want is, or the training expected output, when this block is run/clicked it finds the output of your inputs in that block, then calculates the difference of the output and the expected output using mean-squared-error to find the error. then it goes through each weight, and finds the partial derivative of that weight with respect to the output, and updates according to that and the learning rate. then there's the "run" block, which moves your inputs through the network, through the current weights, to predict the output, the output is shown in the list called "out"
