Machine Learning (EIGHT): Neural Networks (1)

(main reference book "Neural Network and Deep Learning") 1. What is neural network 1.1 from the Perceptron ...

What is a perceptron. Quite simply, as we have said before:

Output=sign (wTx) output=sign (W^TX)
What does that mean. We have some input and we will make a decision based on these inputs: YES OR not. We might think it would be so simple. Then we have to think about computer--cpu and the composition of the human brain.

Start with a simple CPU: simply or not three gate circuits, we can assemble a sufficiently complex CPU. Similarly, as long as we think of the sensor as a unit, after the combination, can simulate the implementation of complex enough functions.
To implement with or by a perceptron:

Suppose x1,x2∈{0,1} x_1,x_2\in\{0,1\},
As long as the w= (−1,1,1) T w= ( -1,1,1) ^t, the output y=sign (−1+x1+x2) y=sign ( -1+x_1+x_2)--This is one with the door;
As long as the w= (−1,2,2) T w= ( -1,2,2) ^t, the output y=sign (−1+2x1+2x2) y=sign ( -1+2x_1+2x_2)--This is the one or the door;

As long as the w= (1,−2) T w= (1,-2) ^t, the output y=sign (1−2x1) y=sign (1-2x_1)--This is a non gate;

with these three basic components, I can use a sensor to assemble an infinitely complex digital circuit . For example, XOR gate: Exclusive (X1,X2) =or (not (x1), x2), and (X1,not (x2)