Your paper and code implementation use real valued spherical harmonics, latter being replete with clean and structured example notebooks (thank you for providing such accessible resources!).
I am curious: was there a specific motivation for selecting real valued spherical harmonics, rather than spin-weighted spherical harmonics, as used in Scaling Spherical CNNs? Was this choice driven by a desire to maintain conceptual simplicity and the associated SO(3) properties that spins can complicate, or were there implementation-related considerations that influenced the decision?
Your paper and code implementation use real valued spherical harmonics, latter being replete with clean and structured example notebooks (thank you for providing such accessible resources!).
I am curious: was there a specific motivation for selecting real valued spherical harmonics, rather than spin-weighted spherical harmonics, as used in Scaling Spherical CNNs? Was this choice driven by a desire to maintain conceptual simplicity and the associated SO(3) properties that spins can complicate, or were there implementation-related considerations that influenced the decision?