ScipyGenerator with scipy.optimize.minimize¶

This notebook demonstrates how to use Xopt's ScipyGenerator to drive any supported scipy optimize.minimize method in a sequential ask/tell workflow.

Imports¶

In [1]:

import matplotlib.pyplot as plt

from xopt import Evaluator, VOCS, Xopt
from xopt.generators.sequential.scipy import ScipyGenerator

import matplotlib.pyplot as plt from xopt import Evaluator, VOCS, Xopt from xopt.generators.sequential.scipy import ScipyGenerator

Define a simple objective¶

We will optimize the 2D Rosenbrock function, exposed through an Xopt evaluator function.

In [2]:

def rosenbrock_eval(input_dict):
    x0 = input_dict["x0"]
    x1 = input_dict["x1"]
    y = (1 - x0) ** 2 + 100.0 * (x1 - x0**2) ** 2
    return {"y": float(y)}

def rosenbrock_eval(input_dict): x0 = input_dict["x0"] x1 = input_dict["x1"] y = (1 - x0) ** 2 + 100.0 * (x1 - x0**2) ** 2 return {"y": float(y)}

Configure ScipyGenerator¶

Choose a scipy method using method. Here we use L-BFGS-B, but methods such as Nelder-Mead, Powell, and others can also be used.

In [3]:

vocs = VOCS(
    variables={"x0": [-2.0, 2.0], "x1": [-1.0, 3.0]},
    objectives={"y": "MINIMIZE"},
)

generator = ScipyGenerator(
    vocs=vocs,
    method="L-BFGS-B",
    initial_point={"x0": -1.2, "x1": 1.0},
    options={"maxiter": 200},
)

evaluator = Evaluator(function=rosenbrock_eval)
X = Xopt(generator=generator, evaluator=evaluator, vocs=vocs)

vocs = VOCS( variables={"x0": [-2.0, 2.0], "x1": [-1.0, 3.0]}, objectives={"y": "MINIMIZE"}, ) generator = ScipyGenerator( vocs=vocs, method="L-BFGS-B", initial_point={"x0": -1.2, "x1": 1.0}, options={"maxiter": 200}, ) evaluator = Evaluator(function=rosenbrock_eval) X = Xopt(generator=generator, evaluator=evaluator, vocs=vocs)

Run optimization¶

In [4]:

for _ in range(200):
    X.step()

best_idx = X.data["y"].argmin()
best = X.data.iloc[best_idx]

print("Evaluations:", len(X.data))
print("Best point:", {"x0": float(best["x0"]), "x1": float(best["x1"])})
print("Best objective:", float(best["y"]))

for _ in range(200): X.step() best_idx = X.data["y"].argmin() best = X.data.iloc[best_idx] print("Evaluations:", len(X.data)) print("Best point:", {"x0": float(best["x0"]), "x1": float(best["x1"])}) print("Best objective:", float(best["y"]))

/home/runner/work/Xopt/Xopt/xopt/generators/sequential/sequential_generator.py:143: RuntimeWarning: scipy minimize has terminated: CONVERGENCE: RELATIVE REDUCTION OF F <= FACTR*EPSMCH. Returning the latest data point; further Xopt.step() calls will re-evaluate it. Check is_done and result, or call reset() to restart.
  candidate = self._generate()

Evaluations: 200
Best point: {'x0': 0.9999969191711192, 'x1': 0.9999937960186889}
Best objective: 9.670715229330395e-12

Inspect convergence¶

In [5]:

ax = X.data["y"].plot(figsize=(7, 4), logy=True)
ax.set_xlabel("iteration")
ax.set_ylabel("objective y (log scale)")
ax.set_title("ScipyGenerator optimization progression")
plt.tight_layout()

ax = X.data["y"].plot(figsize=(7, 4), logy=True) ax.set_xlabel("iteration") ax.set_ylabel("objective y (log scale)") ax.set_title("ScipyGenerator optimization progression") plt.tight_layout()