Why use positional embedding as conditional vector?

[eric-tc]

Hi,
I have read your paper and i do not fully understand why are you using positional embedding as conditional for coupling flow layer. What is the benefit of doing this?

Is related to positional invariance?

Thanks

[gudovskiy]

@eric-tc because positional information allows to incorporate spatial prior to the model. For example, the model with such prior can learn how anomaly features should be arranged on the image.

[mjack3]

My experiments suggest that the model may have performed better without it (green using positional encoding)