READ FOR CONTEXT: This is a tiny neural network that is meant to learn XOR (exclusive OR). This is my first attempt at training a neural network through backpropagation on a small scale. Next I will attempt to build a larger neural network capable of learning to recognise handwritten digits. The goal is that the network outputs 0 if the two binary inputs are the same and 1 if they are different, after training. The target indicates what the output value SHOULD be. When you reset the network or click the green flag, the network is completely random and has not learnt XOR. You can test it by changing the inputs - and the output will most likely be wrong. Then by training the network enough, it should learn how to apply XOR to the inputs. The cost variable should come down over the course of training as that is how incorrect the network is. Sometimes the network fails to learn the correct relationship - in which case it tends to get stuck around 75% accuracy. If this happens, reset the model and train it again. The network is rarely going to output exactly 0 or 1 - more likely a number close to 0 or 1 even after lots of training. This is still considered to be successful. The visuals represent the structure of the network and the activations and weighted connections, shown by the brightness of the nodes/edges
