1. 19 The perceptron can only solve linearly separable problems.

2. A two-layer perceptron can recognise convex areas on its retina.

3. Both models apply the feed forward multilayer perceptron network trained with error back-propagation algorithm.

4. Figure 1 Neural Network as Function Approximator In the next section we will present the multilayer perceptron neural network, and will demonstrate how it can be used as a function Approximator