lsqAxSI 0.0.4

Nonlinear least-squares optimization for AxSI

Usage Guide

Overview

This section describes how to use the Nonlinear Least-Squares Optimization program.

Run from Command Line

To execute the program via the command line, use the following syntax:

lsq_AxSI \
  --x0 1.0 2.0 \
  --bounds -10.0 10.0 -5.0 5.0 \
  --ftol 1e-6 \
  --xtol 1e-6 \
  --diff_step 1e-3 \
  --max_nfev 10000

Required Arguments:

• --x0: Initial guess for the parameters (space-separated values).
• --bounds: Parameter bounds as pairs (lower and upper limits for each parameter).

Optional Arguments:

• --ftol (default: 1e-6): Tolerance for changes in the cost function value.
• --xtol (default: 1e-6): Tolerance for updates to the parameter values.
• --diff_step (default: 1e-3): Step size for finite-difference approximation.
• --max_nfev (default: 10000): Maximum number of function evaluations allowed.

Function Import and Customization

The program also supports direct import of the nonlinear_least_squares function into Python scripts. This allows advanced customization, such as using custom residual (reg_func) and Jacobian (jac) functions.

Example:

from nonlinear_least_squares import nonlinear_least_squares

def custom_reg_func(x, *args):
    # Define custom residual computation
    pass

def custom_jacobian(x, *args):
    # Define custom Jacobian computation
    pass

result = nonlinear_least_squares(
    reg_func=custom_reg_func,
    x0=[1.0, 2.0],
    bounds=([-10.0, -5.0], [10.0, 5.0]),
    jac=custom_jacobian
)

print(result)

Output Description

The program provides a summary of the optimization process, including:

Key Results:

• Success: Indicates whether the optimization converged successfully.
• Optimized Parameters: The final parameter values at the solution.
• Residuals: Values of the residual function at the solution.
• Jacobian: Jacobian matrix at the solution.
• Exit Flag: Integer code indicating the reason for termination.

Detailed Metadata:

The output includes a dictionary with the following fields:

• algorithm: Solver used for optimization.
• firstorderopt: Measure of first-order optimality.
• iterations: Number of iterations performed.
• funcCount: Number of function evaluations.
• cgiterations: Number of conjugate gradient iterations.
• Message: Descriptive message about the termination reason.

Python version

This project is currently using Python 3.12

Installation

It is recommended to use virtualenv to create a clean python environment.

To install lsqAxSI, use pip:

pip install lsqAxSI

Execution

The main script shipped with this project is lsq_AxSI.py, see its options by running:

lsq_AxSI.py -h