Padding & Strides

Can we consider Padding and Strides as hyperparameters which we want to tune to get best filter shape and stride value ?

@Jamal022 , yes, padding and stride are hyper-parameters that you can define. The padding will help to overcome the ‘borders’ effect, which is loosing information from the borders of the input. To overcome this, you can adjust the padding to control this effect. By adding ‘padding’, you will allow the border pixels of the input to ‘participate’ more actively when the filter is applied. You can control this parameter in keras by using the ‘padding’ argument, which by default is zero. As for the Stride, this is also a parameter that you can define and will give you control over the horizontal and vertical applications of your filter. This will have an effect in the size of the resulting features map. In keras this is control with the parameter ‘stride’ which by default is (1,1).