SimplexOptimizer

Implements the Nelder-Mead simplex algorithm for optimization.

Attributes

• objective_function (callable): The objective function to be minimized.

• bounds (list of tuples): The lower and upper bounds for each dimension of the search space.

• n_initial_points (int): The number of initial points to sample.

• n_iter (int): The maximum number of iterations for the optimization process.

• opt_folder (str): The folder to store optimization results.

• max_no_improvement_iters (int): The maximum number of iterations without improvement before stopping.

• converged_threshold (float): The threshold for convergence.

• recorder (LossRecorder): Records and retrieves various metrics throughout the optimization process.

Methods

• optimize(initial_sample_mpi=False, initial_sample_max_processes=1, shrink_mpi=False, shrink_max_process=1): Executes the simplex algorithm to optimize the objective function.

The SimplexOptimizer class provides an interface to perform optimization using the Nelder-Mead simplex algorithm. It initializes with the objective function, search space bounds, and optimization parameters. The optimize method executes the optimization process, optionally sampling initial points and parallelizing operations using MPI.