Plotting and Visualization

All sampler interfaces in discoverysamplers provide consistent plotting methods for visualizing results. The plotting functions are implemented in the plots module and wrapped by each interface.

Available Plots

Method	Description
`plot_trace()`	Trace plots showing parameter evolution over samples
`plot_corner()`	Corner plots with marginal distributions and correlations
`plot_nleaves_histogram()`	Model dimension posterior (RJMCMC only)

Trace Plots

Trace plots show parameter values as a function of sample index, useful for checking convergence and mixing.

Nested Sampling (Nessai, JAX-NS):

# Run sampler
bridge = DiscoveryNessaiBridge(model, priors)
bridge.run_sampler(nlive=1000)

# Plot traces
fig = bridge.plot_trace()
fig.savefig('trace.pdf')

MCMC (Eryn):

For MCMC samplers with parallel tempering, trace plots show all temperatures with different colors:

# Run sampler with parallel tempering
bridge = DiscoveryErynBridge(model, priors)
bridge.create_sampler(nwalkers=32, tempering_kwargs=dict(ntemps=4))
bridge.run_sampler(nsteps=5000)

# Plot traces (discarding burn-in)
fig = bridge.plot_trace(burn=1000)
fig.savefig('trace_mcmc.pdf')

Trace plot for MCMC with parallel tempering

Options:

burn: Number of initial samples to discard
plot_fixed: Include fixed parameters (shown as horizontal lines)

Corner Plots

Corner plots show 1D marginal distributions on the diagonal and 2D projections off-diagonal:

# Basic corner plot
fig = bridge.plot_corner(burn=1000)
fig.savefig('corner.pdf')

# With true values and quantiles
fig = bridge.plot_corner(
    burn=1000,
    truths=[0.5, 1.0, -0.5],      # Mark true values
    quantiles=[0.16, 0.5, 0.84],   # Show 68% CI
)

For MCMC with temperatures, specify which temperature chain to plot:

# Cold chain (temperature 0, the target posterior)
fig = bridge.plot_corner(burn=1000, temp=0)

Model Selection (RJMCMC)

For reversible-jump MCMC, plot the posterior on the number of components:

from discoverysamplers.eryn_RJ_interface import DiscoveryErynRJBridge

# After running RJMCMC
fig = rj_bridge.plot_nleaves_histogram()
fig.savefig('nleaves.pdf')

# With true value marked
from discoverysamplers.plots import plot_nleaves_histogram
nleaves = rj_bridge.return_nleaves()
fig = plot_nleaves_histogram(
    nleaves,
    nleaves_min=1,
    nleaves_max=5,
    true_nleaves=2,
)

Parameter Summary

For a quick overview of parameter estimates with credible intervals:

from discoverysamplers.plots import plot_parameter_summary

samples = bridge.return_sampled_samples()
fig = plot_parameter_summary(samples, credible_interval=0.9)
fig.savefig('summary.pdf')

Using the Plots Module Directly

You can use the plotting functions directly for more control:

from discoverysamplers.plots import plot_trace, plot_corner

# Get samples in standard format
samples = bridge.return_sampled_samples()
# Returns: {'names': [...], 'labels': [...], 'chain': ndarray}

# Create custom plots
fig = plot_trace(
    samples,
    burn=500,
    figsize=(12, 8),
    alpha=0.5,
)

fig = plot_corner(
    samples,
    truths=[0.5, 1.0],
    show_titles=True,
    title_fmt=".2f",
)

Saving Plots

All plotting methods return matplotlib Figure objects:

fig = bridge.plot_corner()

# Save as PDF (vector, publication quality)
fig.savefig('corner.pdf', bbox_inches='tight')

# Save as PNG (raster, for presentations)
fig.savefig('corner.png', dpi=300, bbox_inches='tight')

# Close figure to free memory
import matplotlib.pyplot as plt
plt.close(fig)

Customization with corner

The plot_corner method passes additional keyword arguments to corner.corner():

fig = bridge.plot_corner(
    burn=1000,
    # corner.corner options:
    bins=30,
    smooth=1.0,
    color='C0',
    fill_contours=True,
    levels=[0.68, 0.95],
    plot_datapoints=False,
)

See the corner documentation for all options.