Source code for quantdata.dysanos.tangent_levy.plotting

"""Reproducible diagnostic figures for cached Tangent Lévy artifacts."""

from __future__ import annotations

from pathlib import Path

import numpy as np

from .dynamics import DynamicsFit
from .experiment import HistoricalCalibration


[docs] def write_diagnostic_figure( history: HistoricalCalibration, dynamics: DynamicsFit, *, output: Path, reconstruction_tolerance: float = 0.25, ) -> Path: """Write the core fit, reconstruction, spectrum, and support diagnostics.""" import matplotlib.pyplot as plt output = Path(output) output.parent.mkdir(parents=True, exist_ok=True) daily_rmse = np.asarray( [ np.median([fit.weighted_rmse for fit in day_fit.fits]) for day_fit in history.day_fits ], dtype=np.float64, ) dates = np.asarray(history.dates, dtype="datetime64[D]") figure, axes = plt.subplots(2, 2, figsize=(12, 8), constrained_layout=True) axes[0, 0].plot(dates, daily_rmse, linewidth=0.8) axes[0, 0].set_title("Median maturity fit error") axes[0, 0].set_ylabel("half spreads") axes[0, 0].grid(alpha=0.25) axes[0, 1].plot( dates, history.reconstruction_l1_error, linewidth=0.8, ) axes[0, 1].axhline( reconstruction_tolerance, color="tab:red", linestyle="--", linewidth=0.8, ) axes[0, 1].set_yscale( "symlog", linthresh=max(reconstruction_tolerance, 1.0e-12), ) axes[0, 1].set_title("Cumulative-density projection") axes[0, 1].set_ylabel("relative L1 change") axes[0, 1].grid(alpha=0.25) factor_index = np.arange(1, dynamics.num_factors + 1) axes[1, 0].bar(factor_index, dynamics.explained_variance_ratio) axes[1, 0].set_title("Retained constrained factors") axes[1, 0].set_xlabel("factor") axes[1, 0].set_ylabel("explained variance ratio") axes[1, 0].set_xticks(factor_index) active_mask = ( np.ones(dynamics.beta.shape[1:], dtype=bool) if dynamics.active_mask is None else dynamics.active_mask ) image = axes[1, 1].imshow( active_mask, origin="lower", aspect="auto", extent=( dynamics.jump_x[0], dynamics.jump_x[-1], dynamics.tau[0] * 252.0, dynamics.tau[-1] * 252.0, ), interpolation="nearest", ) axes[1, 1].set_title("Dynamics active support") axes[1, 1].set_xlabel("jump size") axes[1, 1].set_ylabel("business days to maturity") figure.colorbar(image, ax=axes[1, 1], ticks=[0, 1]) figure.savefig(output, dpi=160) plt.close(figure) return output