Numerical Optimization

← Back to Domain-Specific Abstractions by Domain

4 domain-specific abstractions whose origin domain is Numerical Optimization.

• Gauss–Newton Algorithm — An iterative nonlinear least-squares algorithm that linearizes the residual vector and solves the resulting Jacobian least-squares subproblem for each parameter update.
• Golden-section search — A derivative-free interval-reduction algorithm for optimizing a unimodal function by placing interior evaluations in the golden ratio so one point can be reused each iteration.
• Limited-Memory BFGS — A quasi-Newton optimizer that stores a short history of step and gradient-difference pairs and applies the implied inverse-Hessian approximation by two-loop recursion.
• Powell's method — A derivative-free local optimization algorithm that performs successive line minimizations along a changing set of directions and replaces a direction with the net displacement to build approximate conjugacy.