how to define weights to create basic gates

hello i want to know how should I proceed with defining weights while creating neural net.For eg for creating a NOT gate I generally assign a higher negative weight to the variable of which i want to calculate the compliment;x and x0(bias input =1) have weights -30 and 10 respectively .

So the sigmoid function will be g(10 -30x)

where g(x)=1/(1+e^-x)

This is just for one simple gate. How to proceed for complex gates or even the other basic gates such as AND ,NAND etc?....feed forward techniques can be used for complex network of gates but still how to make the basic ones?