Multilayer Feedforward Networks are Universal Approximators

1. Multilayer Feedforward Networks are Universal Approximators, Sonia Hornik 1989;

2. A visual proof that neural nets can compute any function(http://neuralnetworksanddeeplearning.com/chap4.html):

1) Neural nets have a kind of universality;

2) We can only get an approximation that is as good as we wang;

3) The function must be continuous;

3. 神经网络可以以任意精度近似任意连续函数，甚至只有一层隐藏层时.