Value of the learned scale and bias vector?

Hi, did you check the value change of the learned scale and bias vector throughout the training process? I find the value of them change in the first few iterations and remain stable in the rest time on my own classification dataset. I wonder how the learned vectors look like in your paper? Thanks!