Part Number: TDA2PXEVM
Hello
We are transferring a deep learning model onto a TI TDA platform.
Output of this network is a binary classification prediction that is generated by a softmax function.
According to needs of our project, we have to get floating values of these softmax outputs.
For example, [ 0.230, 0.770 ] or [ 0.013, 0.987 ] in which sum of values in each output should be 1
However the softmax we get from this TI platform are exactly 8-bit values which are like this:
[ 255, 240 ]
[ 255, 255 ]
[ 254, 213 ]
[ 221, 254 ]
...
So, the question is, how can we transform these 8-bit values into floating numbers? And what is the codec formula to do this?
Note that these floating values are between 0 and 1 because they are outputs from a softmax function.
