Jolideco Patch Prior Gleam V0.1

Configuration

This is the configuration the deconvolution was run with:

default

MAPDeconvolver
--------------

  n_epochs              : 2000
  beta                  : 1.000
  learning_rate         : 0.100
  compute_error         : False
  stop_early            : False
  stop_early_n_average  : 10
  display_progress      : True
  device                : cpu
  optimizer             : adam

Model Configuration

This is the configuration the Jolideco model:

default

FluxComponents
--------------

  flux                  :

    use_log_flux        : True
    upsampling_factor   : 2
    frozen              : False
    prior               :

      type              : gmm-patches
      stride            : 4
      cycle_spin        : True
      cycle_spin_subpix : False
      jitter            : False
      gmm               :

        type            : gleam-v0.1

      norm              :

        type            : identity

      patch_norm        :

        type            : std-subtract-mean

      device            : cuda:0

Flux

The plot shows the deconvolved flux image, the reference flux and residuals.

Predicted Counts

The plot shows the predicted counts correspoding to the deconvolved flux image, the reference flux and residuals.

Metrics

Metric	Value
MSE	3.515
NMI	1.032
NRMSE	0.719
SSI	0.974

MSE = Mean Squared Error, SSI = Structural Similarity Index, NMI = Normalized Mutual Information, NRMSE = Normalized Root Mean Squared Error

Traces

Files

Results files for download:

spiral3-bg3-jolideco-patch-prior-gleam-v0.1-chandra-result.fits.gz