void kmedians::calculate_median(cluster & current_cluster, point & median) {
const dataset & data = *m_ptr_data;
const std::size_t dimension = data[0].size();
for (size_t index_dimension = 0; index_dimension < dimension; index_dimension++) {
std::sort(current_cluster.begin(), current_cluster.end(),
[this](std::size_t index_object1, std::size_t index_object2)
{
return (*m_ptr_data)[index_object1] > (*m_ptr_data)[index_object2];
});
std::size_t relative_index_median = (std::size_t) (current_cluster.size() - 1) / 2;
std::size_t index_median = current_cluster[relative_index_median];
if (current_cluster.size() % 2 == 0) {
std::size_t index_median_second = current_cluster[relative_index_median + 1];
median[index_dimension] = (data[index_median][index_dimension] + data[index_median_second][index_dimension]) / 2.0;
}
else {
median[index_dimension] = data[index_median][index_dimension];
}
}
}
void agglomerative::calculate_center(const cluster & cluster, point & center) {
const std::vector<point> & data = *m_ptr_data;
const size_t dimension = data[0].size();
center.resize(dimension, 0.0);
for (auto index_point : cluster) {
for (size_t index_dimension = 0; index_dimension < dimension; index_dimension++) {
center[index_dimension] += data[index_point][index_dimension];
}
}
for (size_t index_dimension = 0; index_dimension < dimension; index_dimension++) {
center[index_dimension] /= cluster.size();
}
}
double kmeans::update_center(const cluster & p_cluster, point & p_center) {
point total(p_center.size(), 0.0);
/* for each object in cluster */
for (auto object_index : p_cluster) {
/* for each dimension */
for (size_t dimension = 0; dimension < total.size(); dimension++) {
total[dimension] += (*m_ptr_data)[object_index][dimension];
}
}
/* average for each dimension */
for (size_t dimension = 0; dimension < total.size(); dimension++) {
total[dimension] = total[dimension] / p_cluster.size();
}
const double change = m_metric(p_center, total);
p_center = std::move(total);
return change;
}
