Non-linear Least Squares Fitting in IDL with MPFIT

MPFIT is a port to IDL of the non-linear least squares fitting program MINPACK-1. MPFIT inherits the robustness of the original FORTRAN version of MINPACK-1, but is optimized for performance and convenience in IDL. In addition to the main fitting engine, MPFIT, several specialized functions are provided to fit 1-D curves and 2-D images; 1-D and 2-D peaks; and interactive fitting from the IDL command line. Several constraints can be applied to model parameters, including fixed constraints, simple bounding constraints, and “tying” the value to another parameter. Several data weighting methods are allowed, and the parameter covariance matrix is computed. Extensive diagnostic capabilities are available during the fit, via a call-back subroutine, and after the fit is complete. Several different forms of documentation are provided, including a tutorial, reference pages, and frequently asked questions. The package has been translated to C and Python as well. The full IDL and C packages can be found at http://purl.com/net/mpfit

💡 Research Summary

MPFIT is a comprehensive implementation of the MINPACK‑1 non‑linear least‑squares fitting library adapted for the Interactive Data Language (IDL). The core of MPFIT is the Levenberg‑Marquardt algorithm, the same robust optimization engine that underlies the original FORTRAN MINPACK‑1 code. By re‑engineering the algorithm to exploit IDL’s array handling and memory management, MPFIT delivers high performance while preserving the numerical stability of its predecessor.

The user supplies a model function, a set of observed data, and an initial guess for the parameters. MPFIT then constructs the residual vector (the difference between model predictions and data) and, optionally, the Jacobian matrix either analytically (if the user provides it) or numerically via finite differences. The optimizer iteratively updates the parameter vector to minimize the sum of squared residuals, balancing gradient descent and Gauss‑Newton steps through the Levenberg‑Marquardt damping factor.

A distinctive strength of MPFIT lies in its flexible constraint system. Three constraint types are supported: (1) fixed parameters that remain constant throughout the fit; (2) bounding constraints that enforce upper and/or lower limits on a parameter; and (3) tying constraints that bind one parameter to a function of one or more other parameters. These constraints are incorporated directly into the Jacobian modification step, ensuring that the algorithm respects them without sacrificing convergence properties.

Weighting of data points is equally versatile. By default, MPFIT treats all points equally, but the user can supply per‑point uncertainties σ, causing the residuals to be scaled by 1/σ². Moreover, a user‑defined weighting callback can be employed to implement more sophisticated error models, such as heteroscedastic noise or correlation structures. After convergence, MPFIT computes the parameter covariance matrix using the approximated Hessian (JᵀJ) and scales it by the reduced χ², providing standard errors and correlation coefficients for each fitted parameter.

Diagnostic capabilities are built into both the iterative and post‑fit phases. During each iteration a user‑supplied callback receives the current parameter vector, residuals, χ², iteration count, and other diagnostic information. This enables real‑time monitoring, dynamic adjustment of damping, or early termination if a predefined criterion is met. Upon completion, MPFIT returns a status code (0 for success) together with a detailed report that includes final χ², degrees of freedom, the covariance matrix, and any warning messages (e.g., singular Jacobian, exceeding maximum iterations, or constraint violations).

Beyond the generic fitting engine, MPFIT supplies a suite of specialized wrappers for common scientific tasks. One‑dimensional curve fitting functions simplify fitting analytic forms such as exponentials, Gaussians, or user‑defined models to spectra. Two‑dimensional image fitting utilities (e.g., gauss2d) allow rapid estimation of Gaussian‑shaped sources in astronomical images. Peak‑finding wrappers combine detection algorithms with MPFIT’s optimization to refine peak positions, amplitudes, and widths. An interactive mode lets users click on data points within an IDL display window, adjust parameters on the fly, and instantly visualize the fit, which is especially useful for exploratory data analysis.

Documentation is extensive: a step‑by‑step tutorial walks new users through basic usage, constraint definition, and weighting strategies; reference pages list every function, its arguments, and optional keywords; and a FAQ addresses common pitfalls such as non‑convergence, parameter scaling, and interpretation of error estimates. This thorough documentation lowers the barrier to entry for scientists who may not be experts in numerical optimization.

MPFIT’s algorithmic core has also been ported to C and Python, ensuring that the same fitting logic can be employed in environments outside IDL. This cross‑language consistency aids reproducibility and facilitates integration into larger pipelines that may combine IDL preprocessing with Python‑based analysis or C‑level performance‑critical code. All source code, example scripts, and test data are freely available at http://purl.com/net/mpfit.

In summary, MPFIT brings the proven robustness of MINPACK‑1 to the IDL community while adding modern conveniences such as flexible constraints, sophisticated weighting, real‑time diagnostics, and a rich set of domain‑specific wrappers. Its open‑source nature, thorough documentation, and multi‑language support make it a valuable tool for astronomers, physicists, and engineers who need reliable, high‑precision non‑linear least‑squares fitting for 1‑D spectra, 2‑D images, or any custom model.