RMS formula mistake?

[ejin-photonium]

RMS = $\sqrt{\text{mean}\left(\sum x_i^2\right)}$, not $\text{mean}\left(\sqrt{\sum x_i^2}\right).$ However, it seems that in various places in DeepLens code we use the wrong definition of RMS

example: analysis_spot() in geolens.py

  spot_rms = ((ray_xy_norm**2).sum(-1).sqrt() * ray.is_valid).sum(-1) / (
      ray.is_valid.sum(-1) + EPSILON
  )

I think should be

  spot_rms = torch.sqrt(
      ((ray_xy_norm**2).sum(-1) * ray.is_valid).sum(-1) / (ray.is_valid.sum(-1) + EPSILON)
  )

other places this occur include geolens_pkg/optim.py in autolens, and 2_autolens_rms.py and Autolens_rms tutorial notebook

Pls let me know if you agree and I can submit a PR fixing it. I think the RMS calculation for loss and training is still correct.

[singer-yang]

Thank you! Right here it seems like a mistake. I appreciate your support :-))