Neural Networks are the combination of dozens of perceptron, and each of the perceptron has input and output and the pre-output will be the next input of the other perceptron. Complex combinations of perceptron form the network.
To take an example of the pattern of a single perceptron: I love eating fruits, and I use a function to determine whether I will buy fruits in a supermarket. There are three binary inputs: X1 is whether the supermarket is clean, X2 is whether there are various of choices, X3 is whether the fruits are in good quality. Then I need to weight my options to come up with the best choice, and I’ll add weights to each of the three inputs, three for X1, 2 for X2, 6 for X3, in that way, I got my single perceptron to simply decide whether I'll buy fruits .
![broken image](http://user-assets.sxlcdn.com/images/182442/FpsjeEyyXP6P9b7rK9W143e_3edP.jpg?imageMogr2/strip/auto-orient/thumbnail/1200x9000>/quality/90!/interlace/1/format/jpeg)