Will it limit the receptive field when using convolutional neural network for position encoding?

[AncientRemember]

Intuitively，using cnn for position encoding does have effect！but As we all know, convolutional neural networks are limited by the receptive field，The result is that the position encoding only computed with a small local neighborhoods

[cxxgtxy]

Thanks for your attention. While a very simple PEG has limited receptive field, it is placed after an encoder layer, which already enjoys a global receptive field. The experiment shows that the position of PEG matters. By the way, we have updated the paper to make it more clear.