ScalarDataProblem

class desdeo_problem.problem.ScalarDataProblem(decision_vectors, objective_vectors)[source]

Bases: ProblemBase

A problem class for case where the data is pre-computed.

To be depreciated

Defines a problem with pre-computed data representing a multiobjective optimization problem with scalar valued objective functions.

Parameters:

decision_vectors (np.ndarray) – A 2D vector of decision_vectors. Each row represents a solution with the value for each decision_vectors defined on the columns.
objective_vectors (np.ndarray) – A 2D vector of objective function values. Each row represents one objective vector with the values for the invidual objective functions defined on the columns.

decision_vectors

See args

Type:: np.ndarray

objective_vectors

See args

Type:: np.ndarray

__epsilon

A small floating point number to shift the bounds of the variables. See, get_variable_bounds, default value 1e-6

Type:: float

__constraints

A list of defined constraints.

Type:: List[ScalarConstraint]

nadir

The nadir point of the problem.

Type:: np.ndarray

ideal

The ideal point of the problem.

Type:: np.ndarray

__model_exists

is there a model for this problem

Type:: bool

Note

It is assumed that the decision_vectors and objectives follow a direct one-to-one mapping, i.e., the objective values on the ith row in ‘objectives’ should represent the solution of the multiobjective problem when evaluated with the decision_vectors on the ith row in ‘decision_vectors’.

Attributes Summary

`constraints`	Constraints.
`epsilon`	epsilon.

Methods Summary

`evaluate`(decision_vectors)	Evaluate the values of the objectives at the given decision.
`evaluate_constraint_values`()	Evaluate the constraint values.
`get_variable_bounds`()	Get the variable bounds.

Attributes Documentation

constraints

Constraints.

Returns:: list of the defined constraints
Return type:: List[ScalarConstraint]
Type:: Property

epsilon

epsilon.

Returns:: epsilon value (for shifting the bounds of variables)
Return type:: float
Type:: Property

Methods Documentation

evaluate(decision_vectors)[source]

Evaluate the values of the objectives at the given decision.

Evaluate the values of the objectives corresponding to the decision decision_vectors.

Parameters:

decision_vectors (np.ndarray) – A 2D array with the decision decision_vectors to be evaluated on each row.

Returns:

A 2D array with the objective values corresponding to: each decision vectors on the rows.

Return type:

nd.ndarray

Note

At the moment, this function just maps the given decision decision_vectors to the closest decision variable present (using an L2 distance) in the problem and returns the corresponsing objective vector.

evaluate_constraint_values()[source]

Evaluate the constraint values.

Evaluate the constraint values for each defined constraint. A positive value indicates that a constraint is adhered to, a negative value indicates a violated constraint.

Returns:

A 2D array with each row representing the: constraint values for different objective vectors. One column for each constraint. If no constraint function are defined, returns None.

Return type:

Optional[np.ndarray]

get_variable_bounds()[source]

Get the variable bounds.

Returns:

The variable bounds in a stack. The epsilon value: will be added to the upper bounds and substracted from the lower bounds to return closed bounds.

Return type:

np.array[float]

Note

If self.epsilon is zero, the bounds will represent an open range.