Guyver

Hello, I'm working on a binary classification task. I trained the network and do a forward pass. The result is an output file with the activation outputs. But how do I get the predicted target classes from the network outputs? Is this done by calculating for each output whether it is lower or higer than 0.5? So if output < 0.5 then it is target class A, if output >= 0.5 it is target class B. I tried to do this on the output results, but it didn't give me the same error percentage as I would expect....

Hello, I'm working on a binary classification task. I trained the network and do a forward pass. The result is an output file with the activation outputs. But how do get the predicted target classes from the network outputs? Is this done by calculating for each output whether it is lower or higer than 0.5? So if output < 0.5 then it is target class A, if output >= 0.5 it is target class B. I tried to do this on the output results, but it didn't give me the same error percentage as I would expect....