beluga::clusterizer_detail Namespace Reference

Classes
struct	ClusterCell
	A struct that holds the data of a single cell for the clusterization algorithm. More...

Typedefs
template<class State >
using	ClusterMap = std::unordered_map< std::size_t, ClusterCell< State > >
	A map that holds the sparse data about the particles grouped in cells. More...

Functions
template<class State , class NeighborsFunction >
static void	assign_clusters (ClusterMap< State > &map, NeighborsFunction &&neighbors_function)
	Assign cluster ids to an existing cluster map. More...
template<class Range >
auto	calculate_percentile_threshold (Range &&range, double percentile)
	Calculates the threshold value at a specified percentile from a range. More...
template<class States , class Weights , class Hashes >
static auto	make_cluster_map (States &&states, Weights &&weights, Hashes &&hashes)
	Create a cluster map from a range of particles and their corresponding spatial hashes. More...
template<class Map , class Proj >
auto	make_priority_queue (const Map &map, Proj &&proj)
	Create a priority queue from a map using a specified projection. More...
template<class State >
static void	normalize_and_cap_weights (ClusterMap< State > &map, double percentile)
	Normalize weights and cap them to a given percentile. More...

Typedef Documentation

ClusterMap

template<class State >

using beluga::clusterizer_detail::ClusterMap = typedef std::unordered_map<std::size_t, ClusterCell<State> >

A map that holds the sparse data about the particles grouped in cells.

Definition at line 121 of file cluster_based_estimation.hpp.

Function Documentation

assign_clusters()

template<class State , class NeighborsFunction >

static void beluga::clusterizer_detail::assign_clusters ( ClusterMap< State > &  map, NeighborsFunction &&  neighbors_function  )

static

Assign cluster ids to an existing cluster map.

This function implements a clustering algorithm that assigns cluster IDs to cells in a map based on their spatial relationships.

Notice that with this algorithm each cell will go through the priority queue at most twice.

Template Parameters

State	The state type of the cells in the map.
NeighborsFunction	A callable object that, given a state, returns a range of neighboring cell hashes.

Parameters

map	A reference to the map where cells are stored.
neighbors_function	A function that returns neighboring cell hashes for a given state.

Definition at line 203 of file cluster_based_estimation.hpp.

calculate_percentile_threshold()

template<class Range >

auto beluga::clusterizer_detail::calculate_percentile_threshold	(	Range &&	range,
		double	percentile
	)

Calculates the threshold value at a specified percentile from a range.

Find the value that is greater than the given percentage of the numbers in a range.

Template Parameters

Range

The type of the input range containing the values.

Parameters

range	The input range of values from which to calculate the percentile threshold.
percentile	The percentile (between 0 and 1) to calculate the threshold for.

Returns

The value at the specified percentile in the sorted range.

Definition at line 103 of file cluster_based_estimation.hpp.

make_cluster_map()

template<class States , class Weights , class Hashes >

static auto beluga::clusterizer_detail::make_cluster_map ( States &&  states, Weights &&  weights, Hashes &&  hashes  )

static

Create a cluster map from a range of particles and their corresponding spatial hashes.

This method will populate all the relevant fields in the map except for the cluster ID, which has to be computed with a separate call to assign_clusters.

Template Parameters

States	The range type for particle states.
Weights	The range type for particle weights.
Hashes	The range type for particle spatial hashes.

Parameters

states	A range of particle states.
weights	A range of particle weights.
hashes	A range of particle spatial hashes.

Returns

A map where each unique hash corresponds to a cell with accumulated weight, particle count, and representative state.

Definition at line 138 of file cluster_based_estimation.hpp.

make_priority_queue()

template<class Map , class Proj >

auto beluga::clusterizer_detail::make_priority_queue	(	const Map &	map,
		Proj &&	proj
	)

Create a priority queue from a map using a specified projection.

This function template constructs a priority queue where the elements are ordered by a priority value derived from the map's values. The elements in the queue will contain a key that belongs to the map and the corresponding priority.

Template Parameters

Map	The type of the associative container.
Proj	The type of the projection invocable.

Parameters

map	The map containing the data to be inserted into the priority queue.
proj	The projection function used to compute the priority of each element.

Returns

A priority queue where elements are ordered by the priority computed from the map's values using the projection function.

Definition at line 72 of file cluster_based_estimation.hpp.

normalize_and_cap_weights()

template<class State >

static void beluga::clusterizer_detail::normalize_and_cap_weights ( ClusterMap< State > &  map, double  percentile  )

static

Normalize weights and cap them to a given percentile.

Given a valid cluster map, normalize the accumulated weight by the number of particles in each cell to avoid biasing the clustering algorithm towards cells that randomly end up with more particles than others.

Then cap the values to a given percentile to flatten the top of the approximated density function and make the clustering algorithm more robust to estimation noise, by fusing together any adjacent peaks whose weight is above the threshold.

Template Parameters

State

The state type of the cells in the map.

Parameters

map	A reference to the map where cells are stored.
percentile	The percentile threshold for capping the weights.

Definition at line 174 of file cluster_based_estimation.hpp.