corner.py

2.2.3 verified Mon Apr 27 auth: no python

Corner.py makes beautiful corner plots (scatter plot matrices) for visualizing high-dimensional datasets, commonly used in Bayesian analysis (e.g., MCMC chains). Current version 2.2.3 requires Python >=3.9. Releases are infrequent, mainly maintenance.

pip install corner

Common errors

error ImportError: cannot import name 'corner' from 'corner' ↓

cause Attempting to import `corner` as a function instead of a module.

fix

Use import corner then call corner.corner(samples).

error ValueError: The number of truths (3) does not match the number of dimensions (2) ↓

cause The `truths` list length differs from the number of columns in the samples array.

fix

Ensure truths has the same number of elements as samples.shape[1].

error TypeError: corner() got an unexpected keyword argument 'smooth' ↓

cause Using deprecated `smooth` parameter in corner >=2.2.0.

fix

Use smooth1d and/or smooth2d instead.

error No module named 'arviz' when using 'corner' ↓

cause ArviZ is an optional dependency; some workflows may unexpectedly require it.

fix

Install with pip install corner[arviz] or pip install arviz.

Warnings

gotcha The `corner.corner()` function returns a matplotlib `Figure` object, not a `GridSpec` or an array of axes. Do not try to unpack the return value. ↓

fix Capture the figure: `fig = corner.corner(...)` and use `fig.axes` to access subplots.

deprecated The `smooth` parameter for KDE smoothing is deprecated in favor of `smooth1d` and `smooth2d`. ↓

fix Use `smooth1d=...` and/or `smooth2d=...` instead of `smooth=...`.

gotcha When passing `truths` (true parameter values), the list length must match the dimensionality of the samples (number of columns). ↓

fix Ensure `truths` has the same number of elements as the columns of `samples`.

gotcha The `labels` parameter overrides the `quantiles` display on the histograms. If you want quantiles shown, ensure `labels` is not set or set it to include them. ↓

fix Either omit `labels` or pass a list of label strings. Quantiles will appear if `show_titles=True` and `quantiles` is set.

gotcha Corner plots are memory-intensive for very large datasets (millions of points). Consider thinning your samples before plotting. ↓

fix Subsample your data: `samples = samples[::10]`

Imports

corner
wrong
```
from corner import corner
```
correct
```
import corner
```
corner.corner() is the main function; importing corner.corner directly is not recommended.

Quickstart

Create a simple corner plot from random samples.

import corner
import numpy as np

# Generate random samples
ndim = 5
nsamples = 10000
np.random.seed(42)
samples = np.random.randn(ndim * nsamples).reshape([nsamples, ndim])

# Make the corner plot
figure = corner.corner(samples)

# Save to file
figure.savefig("corner_plot.pdf")