baycomp-plotting 1.2.0

Extra plotting functionality for baycomp's Bayesian classifier comparison posteriors.

Baycomp Plotting

The baycomp_plotting is a python package for building good-looking plots of bayesian posteriors obtained with baycomp.

This package could be useful for scientific purposes, specially in the area of Machine Learning.

Author

• Mario Juez-Gil <[e-mail]>
  Department of Computer Science
  Universidad de Burgos
  ADMIRABLE Research Group

Installation

This package can be installed using PIP.

pip install baycomp_plotting

Basic Usage

The package can be imported as follows:

import baycomp_plotting as bplt

Two plotting functions (tern, and dens), and one class with four matplotlib alternative colors (Color) are provided.

Colors

Four alternative colors to default matplotlib colors are provided:

Example:

import baycomp_plotting as bplt

print(bplt.Color.BLUE)

Output:

'#008ece'

Density plots

For plotting the comparison of two classifiers on a single dataset, dens function could be used. It's parameters are the following:

• p: baycomp posterior.
• label: label of the density function.
• ls: line style (use a matplotlib line style) [default: -]
• color: density function color [default: Color.BLUE]

Example:

import baycomp_plotting as bplt
import baycomp as bc

posterior = bc.CorrelatedTTest(left_classifier_acc, right_classifier_acc, rope=0.01)
fig = bplt.dens(posterior, label='C1', ls='-', color=bplt.Color.BLUE)

Output:

The output figure will have a new function named add_posterior so you can add more posteriors to the figure. The parameters are the same as for dens.

Example:

import baycomp_plotting as bplt
import baycomp as bc

posterior = bc.CorrelatedTTest(left_classifier_1_acc, right_classifier_acc, rope=0.01)
posterior_1 = bc.CorrelatedTTest(left_classifier_2_acc, right_classifier_acc, rope=0.01)
fig = bplt.dens(posterior, label='C1', ls='-', color=bplt.Color.BLUE)
fig.add_posterior(posterior_1, label='C2', ls=(0,(5,1)), color=bplt.Color.GRAY)
fig.legend() # you can show the legend

Output:

Ternary plots

For plotting the comparison of two classifiers on multiple datasets using a ternary plot, tern function could be used. It's parameters are the following:

• p: baycomp posterior.
• names: an array containing Left and Right region labels. [default: ["L", "R"]]

Example:

import baycomp_plotting as bplt
import baycomp as bc

posterior = bc.HierarchicalTest(left_classifier_acc, right_classifier_acc, rope=0.01)
fig = bplt.tern(posterior)

Output:

Comparison against baycomp default plots

Density:

Ternary:

Contribute

Feel free to submit any pull requests 😊

Acknowlegments

This work was supported by the pre-doctoral grant (EDU/1100/2017) of the Consejería de Educación of the Junta de Castilla y León, Spain, and the European Social Fund.

License

This work is licensed under GNU GPL v3.

Citation policy

Please, cite this work as:

@software{baycomp_plotting,
  author       = {Mario Juez-Gil},
  title        = {{mjuez/baycomp_plotting}},
  month        = nov,
  year         = 2020,
  publisher    = {Zenodo},
  version      = {v1.0},
  doi          = {10.5281/zenodo.4244542},
  url          = {https://doi.org/10.5281/zenodo.4244542}
}