Decompose#

class decompose#

The decompose meta-problem.

../../../_images/decompose.png

This meta-problem decomposes a multi-objective input user-defined problem, resulting in a single-objective user-defined problem with a fitness function combining the original fitness functions. In particular, three different decomposition methods are here made available:

  • weighted decomposition,

  • Tchebycheff decomposition,

  • boundary interception method (with penalty constraint).

In the case of \(n\) objectives, we indicate with: \( \mathbf f(\mathbf x) = [f_1(\mathbf x), \ldots, f_n(\mathbf x)] \) the vector containing the original multiple objectives, with: \( \boldsymbol \lambda = (\lambda_1, \ldots, \lambda_n) \) an \(n\)-dimensional weight vector and with: \( \mathbf z^* = (z^*_1, \ldots, z^*_n) \) an \(n\)-dimensional reference point. We also ussume \(\lambda_i > 0, \forall i=1..n\) and \(\sum_i \lambda_i = 1\).

The decomposed problem is thus a single objective optimization problem having the following single objective, according to the decomposition method chosen:

  • weighted decomposition: \( f_d(\mathbf x) = \boldsymbol \lambda \cdot \mathbf f \),

  • Tchebycheff decomposition: \( f_d(\mathbf x) = \max_{1 \leq i \leq m} \lambda_i \vert f_i(\mathbf x) - z^*_i \vert \),

  • boundary interception method (with penalty constraint): \( f_d(\mathbf x) = d_1 + \theta d_2\),

where \(d_1 = (\mathbf f - \mathbf z^*) \cdot \hat {\mathbf i}_{\lambda}\), \(d_2 = \vert (\mathbf f - \mathbf z^*) - d_1 \hat {\mathbf i}_{\lambda})\vert\) and \( \hat {\mathbf i}_{\lambda} = \frac{\boldsymbol \lambda}{\vert \boldsymbol \lambda \vert}\).

See also

“Q. Zhang – MOEA/D: A Multiobjective Evolutionary Algorithm Based on Decomposition” https://en.wikipedia.org/wiki/Multi-objective_optimization#Scalarizing

Note

The reference point \(z^*\) is often taken as the ideal point and as such it may be allowed to change during the course of the optimization / evolution. The argument adapt_ideal activates this behaviour so that whenever a new ideal point is found \(z^*\) is adapted accordingly.

Note

The use of pagmo::decompose discards gradients and hessians so that if the original user defined problem implements them, they will not be available in the decomposed problem. The reason for this behaviour is that the Tchebycheff decomposition is not differentiable. Also, the use of this class was originally intended for derivative-free optimization.

Public Functions

decompose()#

The default constructor will initialize this with a 2-objectives pagmo::null_problem, weight vector [0.5, 0.5] and reference point [0., 0.].

Throws

unspecified – any exception thrown by the other constructor.

template<typename T, ctor_enabler<T> = 0>
inline explicit decompose(T &&p, const vector_double &weight, const vector_double &z, const std::string &method = "weighted", bool adapt_ideal = false)#

Constructor from problem.

Wraps a user-defined problem (UDP) or a pagmo::problem so that its fitness will be decomposed using one of three decomposition methods. pagmo::decompose objects are user-defined problems that can be used to define a pagmo::problem.

Note

This constructor is enabled only if T can be used to construct a pagmo::problem.

Parameters

  • p – the input UDP or pagmo::problem.

  • weight – the vector of weights \(\boldsymbol \lambda\).

  • z – the reference point \(\mathbf z^*\).

  • method – an std::string containing the decomposition method chosen.

  • adapt_ideal – when true, the reference point is adapted at each fitness evaluation to be the ideal point.

Throws

std::invalid_argument – if either:

  • p is single objective or constrained,

  • method is not one of ["weighted", "tchebycheff", "bi"],

  • weight is not of size \(n\),

  • z is not of size \(n\),

  • weight is not such that \(\lambda_i > 0, \forall i=1..n\),

  • weight is not such that \(\sum_i \lambda_i = 1\).

vector_double fitness(const vector_double&) const#

Fitness computation.

The fitness values returned by the inner problem will be combined using the decomposition method selected during construction.

Parameters

x – the decision vector.

Throws

unspecified – any exception thrown by decompose::original_fitness(), or by the fitness decomposition.

Returns

the decomposed fitness of x.

vector_double original_fitness(const vector_double&) const#

Fitness of the original problem.

Returns the fitness of the original multi-objective problem used to construct the decomposed problem.

Note

This is not the fitness of the decomposed problem. Such a fitness is instead returned by calling decompose::fitness().

Parameters

x – input decision vector.

Throws

unspecified – any exception thrown by the original fitness computation.

Returns

the fitness of the original multi-objective problem.

inline vector_double::size_type get_nobj() const#

Number of objectives.

Returns

one.

vector_double::size_type get_nix() const#

Integer dimension.

Returns

the integer dimension of the inner problem.

std::pair<vector_double, vector_double> get_bounds() const#

Box-bounds.

Forwards the bounds computations to the inner pagmo::problem.

Throws

unspecified – any exception thrown by problem::get_bounds().

Returns

the lower and upper bounds for each of the decision vector components.

vector_double get_z() const#

Gets the current reference point.

The reference point to be used for the decomposition. This is only used for Tchebycheff and boundary interception decomposition methods.

Note

The reference point is adapted (and thus may change) at each call of the fitness.

Returns

the reference point.

std::string get_name() const#

Problem name.

This method will append [decomposed] to the name of the inner problem.

Returns

a string containing the problem name.

std::string get_extra_info() const#

Extra information.

This method will add info about the decomposition method to the extra info provided by the inner problem.

Returns

a string containing extra information on the problem.

bool has_set_seed() const#

Calls has_set_seed() of the inner problem.

Calls the method has_set_seed() of the inner problem.

Returns

a flag signalling whether the inner problem is stochastic.

void set_seed(unsigned)#

Calls set_seed() of the inner problem.

Calls the method set_seed() of the inner problem.

Parameters

seed – seed to be set.

Throws

std::not_implemented_error – if the inner problem is not stochastic.

thread_safety get_thread_safety() const#

Problem’s thread safety level.

The thread safety of a meta-problem is defined by the thread safety of the inner pagmo::problem.

Returns

the thread safety level of the inner pagmo::problem.

const problem &get_inner_problem() const#

Getter for the inner problem.

Returns a const reference to the inner pagmo::problem.

Returns

a const reference to the inner pagmo::problem.

problem &get_inner_problem()#

Getter for the inner problem.

Returns a reference to the inner pagmo::problem.

Note

The ability to extract a non const reference is provided only in order to allow to call non-const methods on the internal pagmo::problem instance. Assigning a new pagmo::problem via this reference is undefined behaviour.

Returns

a reference to the inner pagmo::problem.