Repository owner: Jason Christopherson (jchristopherson)
Object-oriented solvers for nonlinear equations, least squares and optimization.
Repository · Pinned source (2133d067886e) · Try in playground · Checked package example
Topics: bfgs, fortran, least-squares, levenberg-marquardt, nelder-mead, newton-raphson-multivariable, newtons-method, nonlinear-equations, optimization, polynomials, quasi-newton, root-finding
Resolved GitHub repositories used by the verified playground build, combining direct and indirect dependencies. This list does not describe internal module dependencies or native system libraries.
Build record verified 2026-10-28.
These selected builds include this repository directly or indirectly. Each link to source identifies the revision that build uses.
These mappings have pinned source evidence. A checked package example may exercise only some of them; source review is not a numerical benchmark.
Module: nonlin
G1b1a: verified
Derivatives / callbacks: none
Accepts a scalar objective through fcnnvar_helper.
Pinned source: src/nonlin_optimize.f90
Author examples/tests: examples/nonlin_nm_opt_example.f90 · examples/example_problems.f90
Module: nonlin
G1b1a: source-supported; G1b1b: verified
Derivatives / callbacks: optional user gradient; finite-difference fallback
User gradients are registered with fcnnvar_helper%set_gradient_fcn. G1b1a applies to finite-difference mode; G1b1b applies when the user supplies the gradient/Jacobian. The driver, where present, identifies its selected configuration.
Pinned source: src/nonlin_optimize.f90 · src/nonlin_multi_var.f90
Author examples/tests: examples/nonlin_bfgs_example.f90 · examples/example_problems.f90
Module: nonlin
G1b1a: source-supported; G1b1b: verified
Derivatives / callbacks: optional user residual Jacobian
The unconstrained least_squares_solver is mapped; constrained_least_squares_solver is a distinct type. For smooth nonlinear least squares, the user residual Jacobian supplies first-derivative information for the squared-residual objective; this is not a general scalar-objective gradient interface. G1b1a applies to finite-difference mode; G1b1b applies when the user supplies the gradient/Jacobian. The driver, where present, identifies its selected configuration.
Pinned source: src/nonlin_least_squares.f90
Author examples/tests: examples/nonlin_least_squares_example.f90
Load checked GAMS driver: gams-nonlin-lsq-jacobian-reference
Module: nonlin
F2: verified
Derivatives / callbacks: optional user Jacobian; finite-difference fallback
Newton solution of nonlinear systems through vecfcn_helper. The checked F2 driver solves [1-x,10*(y-x**2)]=0 from [-1.2,1] and [2,2], requiring the known root [1,1], small recomputed residuals and documented successful termination. Configuration: finite-difference and analytic Jacobians. Other problem dimensions, singular systems and global convergence are not verified.
Pinned source: src/nonlin_solve.f90
Author examples/tests: examples/nonlin_newton_example.f90
Module: nonlin
F1b: verified
Derivatives / callbacks: none
Scalar bracketing root finding; not Brent minimization.
Pinned source: src/nonlin_solve.f90
Author examples/tests: examples/nonlin_brent_example.f90
Module: nonlin
F1b: verified
Derivatives / callbacks: optional first derivative; finite differences otherwise
Safeguarded scalar Newton iteration retaining a bracket. The checked driver supplies an analytic derivative. GAMS F1b has no derivative-specific descendants.
Pinned source: src/nonlin_solve.f90 · src/nonlin_single_var.f90
Author examples/tests: examples/nonlin_newton1d_example.f90