machine learning - How is the desired output of a neural network represented so as to be compared with the actual output? -



machine learning - How is the desired output of a neural network represented so as to be compared with the actual output? -

a neural network learns approximate desired output , hence 1 can utilize unknown testing sets , classify each testing illustration according respective class. illustration neural network can larn classify handwritten digits , recognise handwritten '9' value 9.

how neural network compare output of training sample desired output? encoding desired output take in construction of neural network?

the main thought don't create 1 single output , inquire "what digit this??". create 1 output each digit, , inquire each 1 "is digit x??".

so, desired output must encoded 1xn vector, n number of classes. values 0, , value corresponding desired class 1. in case example, create 1x10 vector, , encode 0 as

[1 0 0 0 0 0 0 0 0 0]

one

[0 1 0 0 0 0 0 0 0 0]

and 9

[0 0 0 0 0 0 0 0 0 1]

for actual output, create perceptron 10 outputs. each output give value in [0,1]. output max value wins, , regarded 1, while others regarded 0. so, if output of network is

[0.1 0.05 0.02 0.92 0.4 ....]

these correspond votes digits [0 1 2 3 4 ..], number suggested classifier three

machine-learning neural-network

Comments

Popular posts from this blog

Delphi change the assembly code of a running process -

json - Hibernate and Jackson (java.lang.IllegalStateException: Cannot call sendError() after the response has been committed) -

C++ 11 "class" keyword -