Dear Author,
Thank you for making your excellent work publicly available.
I have a question regarding the CENet architecture. In the MCA block, you split the tensor along the channel dimension. Could you please explain why this approach was adopted? I understand that Moganet uses a similar strategy, but the DWC layers are organized differently.
Any clarification you could provide would be greatly appreciated.
Best regards,
A struggling PhD student
Dear Author,
Thank you for making your excellent work publicly available.
I have a question regarding the CENet architecture. In the MCA block, you split the tensor along the channel dimension. Could you please explain why this approach was adopted? I understand that Moganet uses a similar strategy, but the DWC layers are organized differently.
Any clarification you could provide would be greatly appreciated.
Best regards,
A struggling PhD student