A Python package for visualizing data with uncertainty using Value-Suppressing Uncertainty Palettes (VSUPs). Inspired by https://github.com/uwdata/vsup.
pip install vsup
The project is developed with uv.
To check for a local python environment, run:
uv run python
Also install the pre-commit hooks with:
uv tool install pre-commit
pre-commit install
from vsup import VSUP
import numpy as np
import matplotlib.pyplot as plt
# Create a grid of values and uncertainties for better visualization
n_points = 50
step = 1 / n_points
values = np.linspace(step / 2, 1 - step / 2, n_points)
uncertainties = np.linspace(step / 2, 1 - step / 2, n_points)
# Create a 2D grid
values, uncertainties = np.meshgrid(values, uncertainties)
# Colorize the data
axs = plt.subplots(3, 3, figsize=(9, 9))[1]
for row, quantization in zip(axs, [None, "linear", "tree"]):
for ax, mode in zip(row, ["us", "ul", "usl"]):
vsup = VSUP(palette="flare", mode=mode, quantization=quantization)
colors = vsup(values, uncertainties)
ax.pcolormesh(values, uncertainties, colors)
# ax.set_title(f"{mode}") #\n({description})")
ax.set_xlabel("Value")
ax.set_ylabel("Uncertainty")
- Three visualization modes:
- USL: Uncertainty mapped to Saturation (chroma) and Lightness
- US: Uncertainty mapped to Saturation
- UL: Uncertainty mapped to Lightness
- Two quantization mods:
- Linear: independent binning of values and uncertainties
- Tree: value bins depend on uncertainty bin: lower uncertainty, higher value resolution
- Support for any matplotlib and seaborn colormaps
If you use this package in your research, please cite the original VSUP paper:
@inproceedings{2018-uncertainty-palettes,
title = {Value-Suppressing Uncertainty Palettes},
author = {Michael Correll AND Dominik Moritz AND Jeffrey Heer},
booktitle = {ACM Human Factors in Computing Systems (CHI)},
year = {2018},
url = {http://idl.cs.washington.edu/papers/uncertainty-palettes},
}
MIT License