Why do we get different predictions when using java and python(dnn)?

asked 2018-02-19 23:02:43 -0500

DenisN03 gravatar image

When using the python language at the output of a neural network, I get an array consisting of numbers from 0 to 5. For example: ...[3 3 3 ... 1 1 1] [3 3 3 ... 1 1 1] [3 3 3 ... 1 1 1]...

When using the same model in java, I get the following: -26.287221908569336, -14.588539123535156, -7.386473655700684, 1.4765703678131104, 15.550891876220703.

Initially, the data comes in a 4d array from which I get values ​​using the reshape function(see this).

edit retag flag offensive close merge delete