How to build a simple neural network in 9 lines of Python code

[Steve D'Aprano]

On Wed, 28 Jun 2017 02:23 am, Sam Chats wrote:

>
https://medium.com/technology-invention-and-more/how-to-build-a-simple-neural-network-in-9-lines-of-python-code-cc8f23647ca1


The derivative of the sigmoid curve given is completely wrong.

    def __sigmoid(self, x):
        return 1 / (1 + exp(-x))

    def __sigmoid_derivative(self, x):
        return x * (1 - x)


Should be:

    def __sigmoid_derivative(self, x):
        return exp(x) / (1 + exp(x))**2



http://mathworld.wolfram.com/SigmoidFunction.html


I wish these neural networks would give a tutorial on how to do something
non-trivial where:

- your training data is not just a couple of bits;

- the function you want the neural network to perform is non-trivial.


E.g. I have a bunch of data:

['cheese', 'pirate', 'aardvark', 'vitamin', 'egg', 'moon']

and a set of flags:

[True, False, True, True, False, True]


and I want the network to predict that:

'calcium' -> False
'aluminium' -> True
'wood' -> True
'plastic' -> False


(The function is, "is there a duplicated vowel?")


-- 
Steve
“Cheer up,” they said, “things could be worse.” So I cheered up, and sure
enough, things got worse.