Visualization¶

qqa.visualization offers consistent plotting helpers that accept a backend="matplotlib" | "plotly" argument. Plotly requires the optional extra pip install qqa[plotly]; missing plotly automatically falls back to matplotlib with a warning.

Catalog¶

from qqa import visualization as viz

viz.plot_history(result)                 # loss / penalty / diversity
viz.plot_best_trajectory(result)         # best objective over epochs
viz.plot_schedule(schedule, num_epochs)  # bg annealing schedule
viz.plot_run_comparison(results, labels=[...])
viz.plot_parallel_coordinates(df, objective="best_obj", backend="plotly")
viz.plot_solution_heatmap(result, problem)

Example: compare multiple runs¶

import qqa
from qqa import visualization as viz

problem = qqa.SherringtonKirkpatrick(N=80, seed=0)
runs = [
    qqa.anneal(problem, num_epochs=600, min_bg=-3, verbose=False),
    qqa.anneal(problem, num_epochs=600, min_bg=-1, verbose=False),
]
viz.plot_run_comparison(runs, labels=["min_bg=-3", "min_bg=-1"])

Integration with dashboards¶

All Plotly figures are raw plotly.graph_objects.Figure instances, so they drop straight into st.plotly_chart(fig).

Gallery¶

Figures below are regenerated deterministically by scripts/make_gallery.py. The full set (eight problems × four plot kinds plus the annealing schedule) lives under data/fig/gallery/ and is referenced from the top-level README. The MkDocs site loads the PNGs directly from GitHub's raw endpoint so the canonical copies stay in data/ without duplication inside docs/.

Schedule¶

Default LinearBGSchedule

Per-problem dynamics¶

Problem	History	Solution	Population
MIS
Max-Cut
Coloring		—
Ising 1D
EA 3D
SK
Perceptron
Hopfield

Regenerate with uv run python scripts/make_gallery.py.