Class prob_dens_mdim_amr (o2scl)

O2scl : Class List

template<class vec_t = std::vector<double>, class mat_t = const_matrix_view_table<vec_t>>
class prob_dens_mdim_amr : public o2scl::prob_dens_mdim<std::vector<double>>

Probability distribution from an adaptive mesh created using a matrix of points.

Todo

Future: The storage required by the mesh is larger than necessary, and could be replaced by a tree-like structure which uses less storage, but that might demand longer lookup times.

Note

This class is experimental.

Dimension choice setting

int dim_choice

Method for choosing dimension to slice.

mutable o2scl::rng rg

Internal random number generator.

bool allow_resampling

If true, allow resampling (default true)

size_t n_dim

Number of dimensions.

vec_t low

Corner of smallest values.

vec_t high

Corner of largest values.

vec_t scale

Vector of length scales.

std::vector<hypercube> mesh

Mesh stored as an array of hypercubes.

int verbose

Verbosity parameter.

static const int max_variance = 1

Choose dimension with maximum variance.

static const int user_scale = 2

Choose dimension with maximum variance with user-specified scale.

static const int random = 3

Choose randomly.

inline virtual int write_generic(std::ostream &fout)

Write the output to a generic output stream.

inline void two_indices_to_density(size_t i, size_t j, table3d &t3d, std::string slice)

Convert two indices to a density in a o2scl::table3d object.

This function presumes that the o2scl::table3d grid has already been created and uses it to create the density. Note that this function will not warn you if the grid refers to points outside the limits of the o2scl::prob_dens_mdim_amr object, instead it will just give zero for those points.

inline void clear()

Clear everything and set the dimensionality to zero.

inline void clear_mesh()

Clear the mesh, leaving the lower and upper limits and the scales unchanged.

inline void copy_to_vectors(size_t &nd, size_t &dc, size_t &ms, std::vector<double> &data, std::vector<size_t> &insides)

Copy the object data to three size_t numbers and two vectors.

Note

This function is used for HDF5 I/O

inline void set_from_vectors(size_t &nd, size_t &dc, size_t &ms, const std::vector<double> &data, const std::vector<size_t> &insides)

Set the object from data specified as three size_t numbers and a set of two vectors.

Note

This function is used for HDF5 I/O

inline void set(vec_t &l, vec_t &h)

Set the mesh limits.

This function is called by the constructor.

Note

Calling this function automatically clears the mesh and the scales.

template<class vec2_t>
inline void set_scale(vec2_t &v)

Set scales for dimension choice.

inline void insert(size_t ir, mat_t &m, bool log_mode = false)

Insert point at row ir, creating a new hypercube for the new point.

inline void initial_parse(mat_t &m, bool log_mode = false)

Parse the matrix m, creating a new hypercube for every point.

inline void initial_parse_new(mat_t &m)

Parse the matrix m, creating a new hypercube for every point, ensuring hypercubes are more optimally arranged.

This algorithm is slower, but may result in more balanced meshes, particularly when dim_choice is not equal to random .

Idea for Future:

This method computes distances twice, once here and once in the insert() function. There is likely a faster approach.

inline void weight_is_inv_volume()

Set the weight in each hypercube equal to the inverse of the volume (the density)

inline double total_volume()

Check the total volume by adding up the fractional part of the volume in each hypercube.

inline double total_weighted_volume()

Check the total volume by adding up the fractional part of the volume in each hypercube.

inline const hypercube &find_hc(const vec_t &x) const

Return a reference to the hypercube containing the specified point.

inline virtual double pdf(const vec_t &x) const

The normalized density.

inline virtual double max_weight() const

Return the maximum weight over all hypercubes in the mesh.

inline virtual double max_frac_vol() const

Return the maximum fractional volume over all hypercubes in the mesh.

inline virtual double max_weighted_vol() const

Return the maximum fractional volume times weight over all hypercubes in the mesh.

inline virtual void select_in_largest(vec_t &x) const

Select a random point in the largest weighted box.

inline virtual void operator()(vec_t &x) const

Sample the distribution.

Public Functions

inline prob_dens_mdim_amr()

Create an empty probability distribution.

inline prob_dens_mdim_amr(vec_t &l, vec_t &h)

Initialize a probability distribution from the corners.

class hypercube

A hypercube class for o2scl::prob_dens_mdim_amr.

Public Functions

inline hypercube()

Create an empty hypercube.

template<class vec2_t>
inline void set(vec2_t &l, vec2_t &h, size_t in, double fvol, double wgt)

Set the hypercube information.

inline hypercube(const hypercube &h)

Copy constructor.

inline hypercube &operator=(const hypercube &h)

Copy constructor through operator=()

template<class vec2_t>
inline bool is_inside(const vec2_t &v) const

Test if point v is inside this hypercube.

Public Members

size_t n_dim

The number of dimensions.

std::vector<double> low

The corner of smallest values.

std::vector<double> high

The corner of largest values.

std::vector<size_t> inside

The list of indices inside.

double frac_vol

The fractional volume enclosed.

double weight

The weight.