参考资料:
在线免费书籍 http://neuralnetworksanddeeplearning.com/chap1.html
Chapter 1
1. perceptron 感知机
- it's a device that makes decisions by weighing up evidence. Just single output.
- inputs 0 or 1(with weights),compared to threshold, then output 0 or 1.
其中,
;
- threshold could be simplified as bias;
其中 b= - threshold, is called bias.
- layers:input layer,hidden layer,output layer
hidden layers(not input and not output)
2. Sigmoid neuron
形式类似于Perceptron, 但是输入输出的值略有变化。这样微小的权重变化,不会引起大的output变化;而Perceptron无法保证。
(small changes Δwj in the weights and Δb in the bias will produce a small change Δoutput in the output from the neuron)
- input : [0,1]
- output:
=
, 其中σ 为sigmoid函数,作用于output.