TS-EMO Python wrappers

The bundled TS-EMO support code includes a few small Python wrappers around BoTorch benchmark functions. They live under qpots/TS-EMO to match the MATLAB implementation layout, so they are shown here as source listings rather than imported with automodule.

Branin-Currin

# Assuming BC_function.py contains the necessary imports and the BraninCurrin class definition

import torch
import numpy as np
from botorch.test_functions import BraninCurrin
from qpots.config import as_tensor

def BC_evaluate(x):
    """
    Evaluate the BoTorch Branin-Currin benchmark for the MATLAB TS-EMO bridge.

    Parameters
    ----------
    x : array-like
        Candidate design points supplied by MATLAB or NumPy with shape
        ``n x 2``.

    Returns
    -------
    numpy.ndarray
        True Branin-Currin objective values evaluated at ``x``.
    """
    # Convert numpy array input to a PyTorch tensor
    X = as_tensor(x)
    
    # Instantiate BraninCurrin problem
    problem = BraninCurrin().to(device=X.device, dtype=X.dtype)
    
    # Evaluate the problem with the given inputs
    result = problem.evaluate_true(X)
    
    # Convert the result back to a numpy array and return
    return result.detach().cpu().numpy()

Car Side Impact

import torch
import numpy as np
from botorch.test_functions.multi_objective import CarSideImpact
from qpots.config import as_tensor

def carside(x):
    """
    Evaluate the BoTorch car-side-impact benchmark.

    Parameters
    ----------
    x : array-like
        Candidate vehicle-design points.

    Returns
    -------
    numpy.ndarray
        Objective values from ``CarSideImpact.evaluate_true``.
    """
    X = as_tensor(x)
    problem = CarSideImpact().to(device=X.device, dtype=X.dtype)
    result = problem.evaluate_true(X)
    return result.detach().cpu().numpy()

DH1

import torch
import numpy as np
from botorch.test_functions.multi_objective import DH1
from qpots.config import as_tensor


def dh1_eval(x, dim):
    """
    Evaluate the DH1 multiobjective benchmark.

    Parameters
    ----------
    x : array-like
        Candidate points with ``dim`` design variables.
    dim : int
        Input dimensionality passed to BoTorch's ``DH1`` constructor.

    Returns
    -------
    numpy.ndarray
        True DH1 objective values.
    """
    X = as_tensor(x)

    problem = DH1(int(dim)).to(device=X.device, dtype=X.dtype)

    result = problem.evaluate_true(X)

    return result.detach().cpu().numpy()

DTLZ1

import torch
import numpy as np
from botorch.test_functions import DTLZ1
from qpots.config import as_tensor

def dtlz1(x, dim):
    """
    Evaluate the two-objective DTLZ1 benchmark.

    Parameters
    ----------
    x : array-like
        Candidate points with ``dim`` design variables.
    dim : int
        Input dimensionality passed to BoTorch's ``DTLZ1`` constructor.

    Returns
    -------
    numpy.ndarray
        True two-objective DTLZ1 values.
    """
    X = as_tensor(x)

    problem = DTLZ1(int(dim), num_objectives=2).to(device=X.device, dtype=X.dtype)
    result = problem.evaluate_true(X)

    return result.detach().cpu().numpy()
    

DTLZ3

import torch
import numpy as np
from botorch.test_functions.multi_objective import DTLZ3
from qpots.config import as_tensor


def dtlz3(x, dim):
    """
    Evaluate the six-objective DTLZ3 benchmark for TS-EMO experiments.

    Parameters
    ----------
    x : array-like
        Candidate points with ``dim`` design variables.
    dim : int
        Input dimensionality passed to BoTorch's ``DTLZ3`` constructor.

    Returns
    -------
    numpy.ndarray
        Negated DTLZ3 objective values. The negation keeps the benchmark
        aligned with the maximization convention used elsewhere in qPOTS.
    """
    X = as_tensor(x)

    problem = DTLZ3(int(dim), num_objectives=6).to(device=X.device, dtype=X.dtype)

    result = problem.evaluate_true(X)
    # DTLZ3 negate=True doesn't work, negate here for results
    return -1 * result.detach().cpu().numpy()

DTLZ7

import torch
import numpy as np
from botorch.test_functions.multi_objective import DTLZ7
from qpots.config import as_tensor


def dtlz7(x, dim):
    """
    Evaluate the two-objective DTLZ7 benchmark.

    Parameters
    ----------
    x : array-like
        Candidate points with ``dim`` design variables.
    dim : int
        Input dimensionality passed to BoTorch's ``DTLZ7`` constructor.

    Returns
    -------
    numpy.ndarray
        True two-objective DTLZ7 values.
    """
    X = as_tensor(x)

    problem = DTLZ7(int(dim), num_objectives=2).to(device=X.device, dtype=X.dtype)
    result = problem.evaluate_true(X)

    return result.detach().cpu().numpy()

Penicillin

import torch
import numpy as np
from botorch.test_functions.multi_objective import Penicillin
from qpots.config import as_tensor


def Penicillin_evaluate(x):
    """
    Evaluate the BoTorch Penicillin simulator benchmark.

    Parameters
    ----------
    x : array-like
        Candidate bioprocess design points.

    Returns
    -------
    torch.Tensor
        True Penicillin benchmark objective values.
    """
    X = as_tensor(x)

    problem = Penicillin().to(device=X.device, dtype=X.dtype)

    result = problem.evaluate_true(X)

    return result

    

Vehicle Safety

import torch
import numpy as np
from botorch.test_functions.multi_objective import VehicleSafety
from qpots.config import as_tensor

def vehiclesafety(x):
    """
    Evaluate the BoTorch vehicle-safety benchmark.

    Parameters
    ----------
    x : array-like
        Candidate vehicle-design points.

    Returns
    -------
    numpy.ndarray
        Objective values from ``VehicleSafety.evaluate_true``.
    """
    X = as_tensor(x)
    problem = VehicleSafety().to(device=X.device, dtype=X.dtype)
    result = problem.evaluate_true(X)
    return result.detach().cpu().numpy()