8.9.2. Plotting Functions¶

plot_residuals(datafile, outfile='default', bins=None, xmax=None, fontsize=22, dpi=200, _verbose=True, **kw)¶

Plots a residuals histogram from any *residuals.dat.

Parameters

datafilestr

Path to *residuals.dat file

outfilestr

Filepath to the .png image to be written. Defaults to datafile.png.

binsint

Number of bins for the histogram plot, Defaults to len(x)/100 or 10, whatever is larger.

xmaxoptional, float > 0

If provided, will set the x-range to (-xmax, xmax).

fontsizefloat

Font size

dpiint

DPI of the outfile

plot_predictions(data, outfile='default', range=None, regplot=True, kde=True, markersize=5, alpha=0.4, fontsize=20, dpi=200, figsize=8, bins=None, rcparams={}, _verbose=True, **kw)¶

Plots the correlation between two distributions x and y. Requires the matplotlib package to be installed.

Parameters

datastr or ScatterXYData

Path to a predictions .txt file generated during a ParAMS optimization. By default located in optimization/training_set_results/best/predictions_*.txt.

Can also be a ScatterXYData, in which case that is plotted.

outfileoptional, str

Filepath to be used for saving. By default (None), will call plt.show() instead – useful for interactive plotting.

rangeoptional, Tuple(float,float)

Allows to change the x and y ranges of the plot (note that xrange=yrange).

regplotbool

If True, will plot a solid line representing the linear regression fit in addition to the single data points.

kdebool

Color-code the data points based on their density using a kernel density estimator. Requires scipy. Disable this plotting is slow, or the number of elements is small (<100).

markersizefloat

Size of the plotted markers.

alpha0 <= float <= 1

Marker transparency. Always 1 if kde`==`True.

fontsizefloat

Font size.

dpiint

DPI of the outfile

binsint > 0

Number of bins to use for the KDE

rcparamsoptional, dict

Optional kwargs dict to be passed to plt.rcParams.

plot_history(datafile, outfile='default', linewidth=2.0, fontsize=20, markers=False, markersize=5, nan_to_zero=False, dpi=120, yscale='linear', _verbose=True, **kw)¶

Plot the cost function values of test / validation sets as a function of total evaluations from a history.dat.

Parameters

datafilestr

Path to history.dat

outfilestr

Filepath to the .png image to be written. Defaults to datafile.png.

linewidthfloat

Line width

fontsizefloat

Font size of the plot

markersbool

Whether to plot markers on top of the lineplot or not.

markersizebool

Marker size, if markers==True

nan_to_zerobool

Whether to convert possible nan / inf in the validation cost values to 0 to guarantee a continuous plot.

dpiint

DPI of the outfile

yscale‘linear’, ‘log’, ‘symlog’, ‘logit’

The pyplot.yscale value

plot_contributions(datafile, outfile='default', cutmin='default', n_max=100, sort=True, print_jobids=True, fontsize=20, dpi=200, _verbose=True, **kw)¶

Generates a barplot based on the contributions from the last evaluate() call, depicting each cost function entry and it’s respective contribution to the total cost function (see Optimization.writefrequency).

Parameters

datafilestr

Path (and fname) where the plot should be stored. Will print the plot on screen (using matplotlib TkAgg) if None.

outfilestr

Filepath to the .png image to be written. Defaults to datafile.png.

cutmin‘default’, float

Entries with a contribution smaller than cutmin will not be plotted. If set to default, will set cutmin = 5 * 100/N_entries (plotting everything that is 5 times larger than the uniform distribution).

n_maxint

Maximum number of entries to show. Will show the entries with the largest contributions first.

sortbool

Whether to sort the entries by descending contributions or not. If None, will sort by the entries’ index in the cost function instead.

print_jobidsbool

Whether to print the jobids of each entry on top of each bar.

fontsizefloat

Font size.

dpiint

DPI of the outfile