Design a 2-input and 1-output multilayer neural network, which could behave as a non-linear classifier to separate the following linearly un-seperable training data. The training data has two attributes and one label. Each attribute could be "1" or "0" and the label is "1" (positive) or "0" (negative). To simplify the problem, every node you use in the multilayer neural network can just perform perceptron function. (List the perceptron function you use).

X1 X2 Output