Plotting

Plotting#

mpol.plot.get_image_cmap_norm(image, stretch='power', gamma=1.0, asinh_a=0.02, symmetric=False)[source]#

Get a colormap normalization to apply to an image.

imagearray

2D image array.

stretchstring, default = ‘power’

Transformation to apply to the colormap. ‘power’ is a power law stretch; ‘asinh’ is an arcsinh stretch.

gammafloat, default = 1.0

Index of power law normalization (see matplotlib.colors.PowerNorm). gamma=1.0 yields a linear colormap.

asinh_afloat, default = 0.02

Scale parameter for an asinh stretch.

symmetricbool, default=False

Whether the colormap is symmetric about 0

mpol.plot.plot_image(image, extent, cmap='inferno', norm=None, ax=None, clab='I [Jy arcsec$^{-2}$]', xlab='$\\Delta \\alpha \\cos \\delta$ [${}^{\\prime\\prime}$]', ylab='$\\Delta \\delta$ [${}^{\\prime\\prime}$]')[source]#

Wrapper for plt.imshow, with colorbar and colormap normalization.

Parameters:

  • image (array) – 2D image array.

  • extent (list, len=4) – x- and y-extents of image: [x-min, x-max, y-min, y-max] (see plt.imshow)

  • cmap (str, default="inferno) – Matplotlib colormap.

  • norm (Matplotlib colormap normalization, default=None) – Image colormap norm. If None, a linear normalization is generated with mpol.plot.get_image_cmap_norm

  • ax (Matplotlib axis instance, default=None) – Axis on which to plot the image. If None, a new figure is created.

  • clab (str, default=r"Jy arcsec$^{-2}$") – Colorbar axis label

  • xlab (str, default="RA offset [arcsec]") – Image x-axis label.

  • ylab (str, default="Dec offset [arcsec]") – Image y-axis label.

Returns:

  • im (Matplotlib imshow instance) – The plotted image.

  • cbar (Matplotlib colorbar instance) – Colorbar for the image.

mpol.plot.vis_histogram_fig(dataset, bin_quantity='count', bin_label=None, q_edges=None, phi_edges=None, q_edges1d=None, show_datapoints=False, save_prefix=None)[source]#

Generate a figure with 1d and 2d histograms of (u,v)-plane coverage. Histograms can show different data; see bin_quantity parameter.

Parameters:

  • dataset (mpol.datasets.GriddedDataset object)

  • bin_quantity (str or numpy.ndarray, default='count') –

    Which quantity to bin:

    • ’count’ bins (u,v) points by their count

    • ’weight’ bins points by the data weight (inherited from dataset)

    • ’vis_real’ bins points by data Re(V)

    • ’vis_imag’ bins points by data Im(V)

    • A user-supplied numpy.ndarray to be used as ‘weights’ in np.histogram

  • bin_label (str, default=None) – Label for 1d histogram y-axis and 2d histogram colorbar.

  • q_edges (array, optional (default=None), unit=:math:[klambda]) – Radial bin edges for the 1d and 2d histogram. If None, defaults to 12 log-linearly radial bins over [0, 1.1 * maximum baseline in dataset].

  • phi_edges (array, optional (default=None), unit=[rad]) – Azimuthal bin edges for the 2d histogram. If None, defaults to 16 bins over [-pi, pi]

  • q_edges1d (array, optional (default=None), unit=:math:[klambda]) – Radial bin edges for a second 1d histogram. If None, defaults to 50 bins equispaced over [0, 1.1 * maximum baseline in dataset].

  • show_datapoints (bool, default = False) – Whether to overplot the raw visibilities in dataset on the 2d histogram.

  • save_prefix (string, default = None) – Prefix for saved figure name. If None, the figure won’t be saved

Returns:

  • fig (Matplotlib .Figure instance) – The generated figure

  • axes (Matplotlib ~.axes.Axes class) – Axes of the generated figure

Notes

No assumption or correction is made concerning whether the (u,v) distances are projected or deprojected.

mpol.plot.split_diagnostics_fig(splitter, channel=0, save_prefix=None)[source]#

Generate a figure showing (u,v) coverage in train and test sets split from a parent dataset.

Parameters:

  • splitter (mpol.crossval.RandomCellSplitGridded object) – Iterator that returns a (train, test) pair of GriddedDataset for each iteration.

  • channel (int, default=0) – Channel (of the datasets in splitter) to use to generate figure

  • save_prefix (string, default = None) – Prefix for saved figure name. If None, the figure won’t be saved

Returns:

  • fig (Matplotlib .Figure instance) – The generated figure

  • axes (Matplotlib ~.axes.Axes class) – Axes of the generated figure

Notes

No assumption or correction is made concerning whether the (u,v) distances are projected or deprojected.

mpol.plot.train_diagnostics_fig(model, losses=None, learn_rates=None, fluxes=None, old_model_image=None, old_model_epoch=None, kfold=None, epoch=None, channel=0, save_prefix=None)[source]#

Figure for model diagnostics at a given model state during an optimization loop.

Plots:

  • model image

  • flux of model image

  • gradient image

  • difference image between old_model_image and current model image

  • loss function

  • learning rate

Parameters:

  • model (torch.nn.Module object) – A neural network module; instance of the mpol.precomposed.GriddedNet class.

  • losses (list) – Loss value at each epoch in the training loop

  • learn_rates (list) – Learning rate at each epoch in the training loop

  • fluxes (list) – Total flux in model image at each epoch in the training loop

  • old_model_image (2D image array, default=None) – Model image of a previous epoch for comparison to current image

  • old_model_epoch (int) – Epoch of old_model_image

  • kfold (int, default=None) – Current cross-validation k-fold

  • epoch (int, default=None) – Current training epoch

  • channel (int, default=0) – Channel (of the datasets in splitter) to use to generate figure

  • save_prefix (str, default = None) – Prefix for saved figure name. If None, the figure won’t be saved

Returns:

  • fig (Matplotlib .Figure instance) – The generated figure

  • axes (Matplotlib ~.axes.Axes class) – Axes of the generated figure

mpol.plot.crossval_diagnostics_fig(cv, title='', save_prefix=None)[source]#

Figure for model diagnostics of a cross-validation run.

Plots:

  • loss evolution for each k-fold

  • cross-validation score per k-fold

Parameters:

  • cv (mpol.crossval.CrossValidate object) – Instance of the CrossValidate class produced by a cross-validation loop

  • title (str, default="") – Figure super-title

  • save_prefix (string, default = None) – Prefix for saved figure name. If None, the figure won’t be saved

Returns:

  • fig (Matplotlib .Figure instance) – The generated figure

  • axes (Matplotlib ~.axes.Axes class) – Axes of the generated figure

mpol.plot.vis_1d_fig(model, uu, vv, V, weights, geom=None, rescale_flux=False, bin_width=20000000.0, q_logx=True, title='', channel=0, save_prefix=None)[source]#
Figure for comparison of 1D projected MPoL model visibilities and observed

visibilities.

Plots:

  • Re(V): observed and MPoL model (projected unless geom is supplied)

  • Residual Re(V): observed - MPoL model (projected unless geom is supplied)

  • Im(V): observed and MPoL model (projected unless geom is supplied)

  • Residual Im(V): observed - MPoL model (projected unless geom is supplied)

Parameters:

  • model (torch.nn.Module object) – A neural network; instance of the mpol.precomposed.GriddedNet class.

  • uu (array, unit=[\(\lambda\)]) – Data u- and v-coordinates

  • vv (array, unit=[\(\lambda\)]) – Data u- and v-coordinates

  • V (array, unit=[Jy]) – Data visibility amplitudes

  • weights (array, unit=[Jy^-2]) – Data weights

  • geom (dict) –

    Dictionary of source geometry. If passed in, visibilities will be
    deprojected prior to plotting. Keys:
    ”incl”float, unit=[deg]

    Inclination

    ”Omega”float, unit=[deg]

    Position angle of the ascending node

    ”omega”float, unit=[deg]

    Argument of periastron

    ”dRA”float, unit=[arcsec]

    Phase center offset in right ascension. Positive is west of north.

    ”dDec”float, unit=[arcsec]

    Phase center offset in declination.

  • rescale_flux (bool) –

    If True, the visibility amplitudes are rescaled to account

    for the difference between the inclined (observed) brightness and the assumed face-on brightness, assuming the emission is optically thick. The source’s integrated (2D) flux is assumed to be: \(F = \cos(i) \int_r^{r=R}{I(r) 2 \pi r dr}\). No rescaling would be appropriate in the optically thin limit.

  • bin_width (float, default=20e3) – Bin size [\(\lambda\)] for baselines

  • q_logx (bool, default=True) – Whether to plot visibilities in log-baseline

  • title (str, default="") – Figure super-title

  • channel (int, default=0) – Channel of the model to use to generate figure

  • save_prefix (string, default = None) – Prefix for saved figure name. If None, the figure won’t be saved

Returns:

  • fig (Matplotlib .Figure instance) – The generated figure

  • axes (Matplotlib ~.axes.Axes class) – Axes of the generated figure

Notes

This routine requires the frank package

mpol.plot.radial_fig(model, geom, u=None, v=None, V=None, weights=None, dist=None, rescale_flux=False, bin_width=20000.0, q_logx=True, title='', channel=0, save_prefix=None)[source]#

Figure for analysis of 1D (radial) brightness profile of MPoL model image, using a user-supplied geometry.

Plots:

  • MPoL model image

  • 1D (radial) brightness profile extracted from MPoL image (supply dist to show second x-axis in [AU])

  • Deprojectd Re(V): binned MPoL model and observations (if u, v, V, weights supplied)

  • Deprojected Im(V): binned MPoL model and observations (if u, v, V, weights supplied)

Parameters:

  • model (torch.nn.Module object) – A neural network; instance of the mpol.precomposed.GriddedNet class.

  • geom (dict) –

    Dictionary of source geometry. Used to deproject image and visibilities.
    Keys:
    ”incl”float, unit=[deg]

    Inclination

    ”Omega”float, unit=[deg]

    Position angle of the ascending node

    ”omega”float, unit=[deg]

    Argument of periastron

    ”dRA”float, unit=[arcsec]

    Phase center offset in right ascension. Positive is west of north.

    ”dDec”float, unit=[arcsec]

    Phase center offset in declination.

  • u (array, optional, unit=[\(\lambda\)], default=None) – Data u- and v-coordinates

  • v (array, optional, unit=[\(\lambda\)], default=None) – Data u- and v-coordinates

  • V (array, optional, unit=[Jy], default=None) – Data visibility amplitudes

  • weights (array, optional, unit=[Jy^-2], default=None) – Data weights

  • dist (float, optional, unit = [AU], default = None) – Distance to source, used to show second x-axis for I(r) in [AU]

  • rescale_flux (bool) – If True, the visibility amplitudes are rescaled to account for the difference between the inclined (observed) brightness and the assumed face-on brightness, assuming the emission is optically thick. The source’s integrated (2D) flux is assumed to be: \(F = \cos(i) \int_r^{r=R}{I(r) 2 \pi r dr}\). No rescaling would be appropriate in the optically thin limit.

  • bin_width (float, default=20e3) – Bin size [klambda] in which to bin observed visibility points

  • q_logx (bool, default=True) – Whether to plot visibilities in log-baseline

  • title (str, default="") – Figure super-title

  • channel (int, default=0) – Channel of the model to use to generate figure

  • save_prefix (string, default = None) – Prefix for saved figure name. If None, the figure won’t be saved

Returns:

  • fig (Matplotlib .Figure instance) – The generated figure

  • axes (Matplotlib ~.axes.Axes class) – Axes of the generated figure

Notes

This routine requires the frank package