Directsearch Optimization Solver

1.0.1 · active · verified Wed Apr 15

directsearch is a Python package for solving unconstrained minimization problems without requiring derivatives of the objective function. It is particularly useful when objective function evaluations are expensive or noisy, implementing a family of direct search methods. The current version is 1.0.1, with releases focused on stability and compatibility.

Warnings

gotcha Older versions of directsearch (e.g., v1.0.0) may encounter errors or deprecation warnings with newer NumPy versions due to internal API changes in NumPy. Version 1.0.1 specifically addresses this.
Fix: Upgrade to directsearch v1.0.1 or newer: `pip install --upgrade directsearch`.
gotcha The initial guess `x0` passed to `directsearch.solve` must be a one-dimensional `numpy.ndarray`.
Fix: Ensure `x0` is created as `np.array([...])` and `len(x0.shape) == 1`.
gotcha The objective function `f` passed to `directsearch.solve` must return a single `float` value.
Fix: Verify that your objective function's return type is a scalar float.

Install

pip install directsearch Install latest version

Imports

solve
```
from directsearch import solve
```
The primary function for initiating a direct search optimization.

Quickstart

This example demonstrates how to use `directsearch.solve` to minimize the Rosenbrock function. The objective function `f` must take a one-dimensional NumPy array and return a single float. The initial guess `x0` must also be a one-dimensional NumPy array.

import numpy as np
from directsearch import solve

def rosenbrock(x):
    # The Rosenbrock function (a common optimization test problem)
    return (1.0 - x[0])**2 + 100.0 * (x[1] - x[0]**2)**2

# Initial guess for the minimizer
x0 = np.array([0.0, 0.0])

# Solve the optimization problem
soln = solve(rosenbrock, x0)

print(f"Optimal value: {soln.f}")
print(f"Optimal solution: {soln.x}")

view raw JSON →