strange accuracy

[kzamzam]

I am running the network on my dataset. I am getting some strange values that I am not sure I fully understand. While training the dataset, in the validation part, I get the baseline log squared norm as 0.00715, while the corrected is 0.00323. The baseline mean Norm is 0.28344, while the corrected is 17.9 which is weird since the log squared norm decreases while this increases, why is that? I think there might be an error with the values in the .mat file. Please please please, anyone help. I will attach a photo of the terminal

.