ScientNLS

scientNLS is an open-source software library for nonlinear least squares (NLS) optimization designed for scientific data analysis and parameter estimation. It provides solvers for unconstrained and constrained problems, including variants of the Levenberg–Marquardt algorithm and trust-region methods, with a focus on numerical stability and performance on large-scale problems.

Development of scientNLS began in 2020 as a collaborative effort among researchers in computational science. The project is hosted on a public repository and released under the MIT license. It aims to be portable across Linux, Windows, and macOS, with bindings to popular languages.

Key features include support for automatic differentiation, sparse Jacobians, multiple solver backends, and robust convergence monitoring. The library emphasizes ease of use through a high-level API and optional performance optimizations such as multi-threading and, where applicable, GPU acceleration.

The core is written in C++ with interfaces to Python and Julia. It uses the Eigen library for linear algebra and provides a modular cost-function framework that allows users to plug in custom models and data sets.

ScientNLS is used by researchers in engineering, physics, and life sciences for tasks such as curve fitting, parameter estimation in differential equations, and model calibration. Documentation, tutorials, and example data accompany the release to facilitate adoption.

As an active project, scientNLS receives ongoing contributions from its user community and is referenced in several scholarly works that rely on nonlinear least-squares methods for data analysis.