The goal of backpropagation is to compute the partial derivatives b in the network. 

 we use the quadratic cost function

   review backpropagation

 

two assumptions :

  1: The first assumption we need is that the cost function can be written as an average 

    review backpropagation    (case for the quadratic cost function)

    The reason we need this assumption is because what backpropagation actually lets us do is compute the partial derivatives

 by averaging over training examples. In

 subscript, writing the

 back in, but for now it's a notational nuisance that is better left implicit.

 

  2: The cost function can be written as a function of the outputs from the neural network

   review backpropagation

the Hadamard product

   (s⊙t)j=sjtj

  review backpropagation

The four fundamental equations behind backpropagation

   review backpropagation

 

 

 

BP1 

   review backpropagation:the error in the jth neuron in the lth layer

     review backpropagation

    You might wonder why the demon is changing the weighted input . Surely it'd be more natural to imagine the demon changing

 as our measure of error. In fact, if you do this things work out quite

  similarly to the discussion below. But it turns out to make the presentation of backpropagation a little more algebraically complicated.

 as our measure of error.

   An equation for the error in the output layer,  are given by

  review backpropagation

  it's easy to rewrite the equation in a matrix-based form, as

  review backpropagation

  review backpropagation

  review backpropagation

BP2

  review backpropagation

  

  review backpropagation

 

BP3

  review backpropagation

  review backpropagation

BP4

  review backpropagation

  review backpropagation

  review backpropagation

 

The backpropagation algorithm

  review backpropagation

 

    review backpropagation

 

      Of course, to implement stochastic gradient descent in practice you also need an outer loop generating mini-batches

    of training examples, and an outer loop stepping through multiple epochs of training. I've omitted those for simplicity.

 

 reference: http://neuralnetworksanddeeplearning.com/chap2.html

------------------------------------------------------------------------------------------------

review backpropagation

review backpropagation

 

reference: Machine Learning by Andrew Ng

相关文章: