Quickstart¶

import numpy as np
import matplotlib.pyplot as plt
%config InlineBackend.figure_format = 'svg'

rng = np.random.default_rng(0)
N = 20_000

data = {
    "mass_1": rng.normal(30, 4, N),
    "mass_2": rng.normal(25, 3, N),
    "chi":    rng.uniform(-1, 1, N),
}

import numpy as np import matplotlib.pyplot as plt %config InlineBackend.figure_format = 'svg' rng = np.random.default_rng(0) N = 20_000 data = { "mass_1": rng.normal(30, 4, N), "mass_2": rng.normal(25, 3, N), "chi": rng.uniform(-1, 1, N), }

First plot¶

corner() takes a dict[str, array] and returns a standard (fig, axes) pair.

from cornetto import corner

fig, axes = corner(data, chain_labels=["Chain 1"])
plt.show()

from cornetto import corner fig, axes = corner(data, chain_labels=["Chain 1"]) plt.show()

Labels and truth markers¶

Pass labels as a dict[str, str] - LaTeX via raw strings works out of the box. Pass truths as a dict[str, scalar] for crosshairs.

labels = {
    "mass_1": r"$m_1\,[M_\odot]$",
    "mass_2": r"$m_2\,[M_\odot]$",
    "chi":    r"$\chi_{\mathrm{eff}}$",
}

fig, axes = corner(
    data,
    labels=labels,
    truths={"mass_1": 30.0, "mass_2": 25., "chi": 0.0},
    chain_labels=["Chain 1"],
)
plt.show()

labels = { "mass_1": r"$m_1\,[M_\odot]$", "mass_2": r"$m_2\,[M_\odot]$", "chi": r"$\chi_{\mathrm{eff}}$", } fig, axes = corner( data, labels=labels, truths={"mass_1": 30.0, "mass_2": 25., "chi": 0.0}, chain_labels=["Chain 1"], ) plt.show()

Crosshairs properties can be set via the parameter kwargs_truths, a dictionary with: color, lw, ls, alpha, marker, and markersize.

Two-chain comparison¶

Pass a 2-D (N_chains, N_samples) array for each parameter and chain_labels to name them in the legend.

data_2 = {
    "mass_1": np.stack([rng.normal(30, 4, N), rng.normal(85, 8, N)]),
    "mass_2": np.stack([rng.normal(25, 3, N), rng.normal(66, 6, N)]),
    "chi":    np.stack([rng.uniform(-1, 1, N), rng.uniform(-1, 1, N)]),
}

labels_2 = {
    "mass_1": r"$m_1\,[M_\odot]$",
    "mass_2": r"$m_2\,[M_\odot]$",
    "chi":    r"$\chi_{\mathrm{eff}}$",
}

fig, axes = corner(
    data_2,
    labels=labels_2,
    truths={"mass_1": np.array([30.0, 85.0]), "mass_2": np.array([25., 66.])},
    chain_labels=["Chain 1", "Chain 2"],
)
plt.show()

data_2 = { "mass_1": np.stack([rng.normal(30, 4, N), rng.normal(85, 8, N)]), "mass_2": np.stack([rng.normal(25, 3, N), rng.normal(66, 6, N)]), "chi": np.stack([rng.uniform(-1, 1, N), rng.uniform(-1, 1, N)]), } labels_2 = { "mass_1": r"$m_1\,[M_\odot]$", "mass_2": r"$m_2\,[M_\odot]$", "chi": r"$\chi_{\mathrm{eff}}$", } fig, axes = corner( data_2, labels=labels_2, truths={"mass_1": np.array([30.0, 85.0]), "mass_2": np.array([25., 66.])}, chain_labels=["Chain 1", "Chain 2"], ) plt.show()

Fast plotting - quick_corner¶

When you just need a sanity check, quick_corner uses histograms instead of KDE - same API, ~7× faster.

from cornetto import quick_corner

fig, axes = quick_corner(data, chain_labels=["Chain 1"])
plt.show()

from cornetto import quick_corner fig, axes = quick_corner(data, chain_labels=["Chain 1"]) plt.show()

The Cornetto class¶

Use Cornetto when you want to render multiple figures from the same data - KDEs are computed once and cached.

from cornetto import Cornetto

c = Cornetto(data, labels=labels, chain_labels=["Chain 1"])
fig, axes = c.plot(color="coral", dark=True)
plt.show()

c.info()      # prints the exact setup used
c.summary()   # summary table (HTML in Jupyter, plain text in terminal)

from cornetto import Cornetto c = Cornetto(data, labels=labels, chain_labels=["Chain 1"]) fig, axes = c.plot(color="coral", dark=True) plt.show() c.info() # prints the exact setup used c.summary() # summary table (HTML in Jupyter, plain text in terminal)

Cornetto setup
--------------
  params         : ['mass_1', 'mass_2', 'chi']
  n_chains       : 1
  chain_labels   : ['Chain 1']
  stat           : median  (center line + shaded interval on 1-D marginals)
  contour sigmas : 1σ→39.3%, 2σ→86.5%
  truths         : none
  ── last plot() ──
  color        : coral
  theme          : dark
  smooth         : True  (KDE if True, histogram if False)
  n_grid         : 128
  bandwidth      : Silverman × 0.8
  fill_1d        : False
  fill_contours  : True

param	chain	median	p16	p84	mean	std
mass_1	Chain 1	29.98	26.08	34	30.02	3.984
mass_2	Chain 1	25	21.96	27.98	25	3.023
chi	Chain 1	-0.007335	-0.683	0.6804	-0.001651	0.5783

Saving¶

PDF and SVG preserve vector quality for papers and talks.

fig.savefig("posterior.pdf", bbox_inches="tight")
fig.savefig("posterior.svg", bbox_inches="tight")

fig.savefig("posterior.pdf", bbox_inches="tight") fig.savefig("posterior.svg", bbox_inches="tight")

Next steps¶

• User guide — weights, statistics, styling, pairplot.
• Examples gallery — marginals, traces, advanced styling.
• Fast mode — quick_corner in depth.
• API reference — every parameter documented.