*
* @license AGPL-3.0-or-later
*
* This code is free software: you can redistribute it and/or modify
* it under the terms of the GNU Affero General Public License, version 3,
* as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU Affero General Public License for more details.
*
* You should have received a copy of the GNU Affero General Public License, version 3,
* along with this program. If not, see
*
*/
namespace OCA\FaceRecognition\Clusterer;
/**
* This class implements the graph clustering algorithm described in the
* paper: Chinese Whispers - an Efficient Graph Clustering Algorithm and its
* Application to Natural Language Processing Problems by Chris Biemann.
*
* In particular, it tries to be a shameless copy of the original dlib
* implementation.
* - https://github.com/davisking/dlib/blob/master/dlib/clustering/chinese_whispers.h
*/
class ChineseWhispers {
/**
* Cluster the dataset by assigning a label to each sample.from the edges
*/
static public function predict(array &$edges, array &$labels, int $num_iterations = 100)
{
// To improve the stability of the clusters, we must
// iterate the neighbors in a pseudo-random way.
mt_srand(2023);
$labels = [];
if (count($edges) == 0)
return 0;
$neighbors = [];
self::find_neighbor_ranges($edges, $neighbors);
// Initialize the labels, each node gets a different label.
for ($i = 0; $i < count($neighbors); ++$i)
$labels[$i] = $i;
for ($iter = 0; $iter < count($neighbors)*$num_iterations; ++$iter)
{
// Pick a random node.
$idx = mt_rand()%count($neighbors);
// Count how many times each label happens amongst our neighbors.
$labels_to_counts = [];
$end = $neighbors[$idx][1];
for ($i = $neighbors[$idx][0]; $i != $end; ++$i)
{
$iLabelFirst = $edges[$i][1];
$iLabel = $labels[$iLabelFirst];
if (isset($labels_to_counts[$iLabel]))
$labels_to_counts[$iLabel]++;
else
$labels_to_counts[$iLabel] = 1;
}
// find the most common label
// std::map::iterator i;
$best_score = PHP_INT_MIN;
$best_label = $labels[$idx];
foreach ($labels_to_counts as $key => $value)
{
if ($value > $best_score)
{
$best_score = $value;
$best_label = $key;
}
}
$labels[$idx] = $best_label;
}
// Remap the labels into a contiguous range. First we find the
// mapping.
$label_remap = [];
for ($i = 0; $i < count($labels); ++$i)
{
$next_id = count($label_remap);
if (!isset($label_remap[$labels[$i]]))
$label_remap[$labels[$i]] = $next_id;
}
// now apply the mapping to all the labels.
for ($i = 0; $i < count($labels); ++$i)
{
$labels[$i] = $label_remap[$labels[$i]];
}
return count($label_remap);
}
static function find_neighbor_ranges (&$edges, &$neighbors) {
// setup neighbors so that [neighbors[i].first, neighbors[i].second) is the range
// within edges that contains all node i's edges.
$num_nodes = self::max_index_plus_one($edges);
for ($i = 0; $i < $num_nodes; ++$i) $neighbors[$i] = [0, 0];
$cur_node = 0;
$start_idx = 0;
for ($i = 0; $i < count($edges); ++$i)
{
if ($edges[$i][0] != $cur_node)
{
$neighbors[$cur_node] = [$start_idx, $i];
$start_idx = $i;
$cur_node = $edges[$i][0];
}
}
if (count($neighbors) !== 0)
$neighbors[$cur_node] = [$start_idx, count($edges)];
}
static function max_index_plus_one ($pairs): int {
if (count($pairs) === 0)
{
return 0;
}
else {
$max_idx = 0;
for ($i = 0; $i < count($pairs); ++$i)
{
if ($pairs[$i][0] > $max_idx)
$max_idx = $pairs[$i][0];
if ($pairs[$i][1] > $max_idx)
$max_idx = $pairs[$i][1];
}
return $max_idx + 1;
}
}
static function convert_unordered_to_ordered (&$edges, &$out_edges)
{
$out_edges = [];
for ($i = 0; $i < count($edges); ++$i)
{
$out_edges[] = [$edges[$i][0], $edges[$i][1]];
if ($edges[$i][0] != $edges[$i][1])
$out_edges[] = [$edges[$i][1], $edges[$i][0]];
}
}
}