The most important thing we should make clear is the map in this neural networks model.Here’s the model->
There’s no doubt that the image above is a quite complex map.Dizzying lines make us disgusting.
Thus, how to describle these lines simplely is an urgent problem.Luckily, there has already been a way put forward by someone.
Let’s look at an example by Andrew Ng ->
Explanation
a subscript i supscript i means activation
of unit i in layer j.
theta supscript j means matrix weights controlling function mapping from layer j to layer j+1.
What’s more, theta supscript j here is a matrix with s subscript j+1
* (s subscript j
+1), e.g. from the image above, we can find that theta supscript 1 here is a 3 by 4 matrix.
in the last 2 videos, Andrew Ng gave us some examles to help us understand this pattern better ->
From these,you will find that we can use these models to easily stimulate almost all logical operations like AND,OR,XOR. Of course, what needs to pay attention is sometimes(often), we add a +1
as a bias to help us stimulate this process better.