pulsatrix 1.1.0
Loading...
Searching...
No Matches
pulsatrix::DistanceOptions Struct Reference

PairwiseDistances()'s settings. More...

#include <distance_matrix.hpp>

Public Attributes

DistanceMethod method = DistanceMethod::Gemm
 How to compute the distances.
 
bool center = false
 Subtract the mean point first. It doesn't change Euclidean distances, but it shrinks the norms, and with them the Gemm method's error, when the points share a large offset (an anisotropic layer).
 
bool normalize = false
 Scale every point to unit length (after centering, if centering). Distances are then sqrt(2 - 2 cos), a metric that orders pairs as cosine similarity does.
 
int64_t block_rows = 0
 Rows per GEMM block; 0 picks one so a block's output holds at most 2^24 values. The device holds the points and one block at a time.
 

Detailed Description

PairwiseDistances()'s settings.

Member Data Documentation

◆ block_rows

int64_t pulsatrix::DistanceOptions::block_rows = 0

Rows per GEMM block; 0 picks one so a block's output holds at most 2^24 values. The device holds the points and one block at a time.

◆ center

bool pulsatrix::DistanceOptions::center = false

Subtract the mean point first. It doesn't change Euclidean distances, but it shrinks the norms, and with them the Gemm method's error, when the points share a large offset (an anisotropic layer).

◆ method

DistanceMethod pulsatrix::DistanceOptions::method = DistanceMethod::Gemm

How to compute the distances.

◆ normalize

bool pulsatrix::DistanceOptions::normalize = false

Scale every point to unit length (after centering, if centering). Distances are then sqrt(2 - 2 cos), a metric that orders pairs as cosine similarity does.


The documentation for this struct was generated from the following file: