nlsq.constants module

Constants for NLSQ optimization algorithms.

These values are derived from: - SciPy’s trust-region algorithms - Numerical optimization best practices - JAX performance characteristics

Overview

The constants module defines numerical constants and thresholds used throughout NLSQ.

Constants

This module provides:

  • Machine epsilon values for different dtypes

  • Convergence tolerances for optimization algorithms

  • Numerical stability thresholds

  • Default parameter values

Example Usage

from nlsq.constants import MACHINE_EPSILON, DEFAULT_XTOL, DEFAULT_FTOL

# Use in convergence checks
if abs(f_new - f_old) < DEFAULT_FTOL:
    print("Converged based on function tolerance")

# Machine precision aware computations
safe_denominator = max(value, MACHINE_EPSILON)

See Also