"Hands-on learning deep learning" learning record 02: Chapter03_2: softmax regression

model:

　　

　　

Softmax:

　　If the direct use of the output layer of two problems result:

　　　　1, since the range of the output value of the output layer is uncertain, it is difficult to intuitively we determined the meaning of these values

　　　　2, since the tag is a discrete real values, the error between the discrete values ​​and the output values ​​of the uncertainty range is difficult to measure

　　By the following formula is converted into an output value and a positive value and a probability distribution:

　　

　　and,

　　

　　The final expression vector as above:

　　

Cross entropy loss function: