HMLP: High-performance Machine Learning Primitives
|
This is a helper class that parses the arguments from command lines. More...
#include <gofmm.hpp>
Public Member Functions | |
CommandLineHelper (int argc, char *argv[]) | |
Public Attributes | |
size_t | n |
size_t | m |
size_t | k |
size_t | s |
size_t | nrhs |
double | stol = 1E-3 |
double | budget = 0.0 |
DistanceMetric | metric = ANGLE_DISTANCE |
size_t | d |
size_t | nb |
double | h = 1.0 |
string | distance_type |
string | spdmatrix_type |
string | kernelmatrix_type |
string | hidden_layers |
string | user_matrix_filename |
string | user_points_filename |
This is a helper class that parses the arguments from command lines.
|
inline |
(Default) constructor.
Number of columns and rows, i.e. problem size.
On-diagonal block size, such that the tree has log(n/m) levels.
Number of neighbors to use.
Maximum off-diagonal ranks.
Number of right-hand sides.
Desired approximation accuracy.
The maximum percentage of direct matrix-multiplication.
Specify distance type.
Specify what kind of spdmatrix is used.
NOP
NOP
NOP
(Optional) provide the path to the matrix file.
(Optional) provide the path to the data file.
Dimension of the data set.
Number of attributes (dimensions).
Number of attributes (dimensions)
Block size (in dimensions) per file
Number of attributes (dimensions)
(Optional) provide Gaussian kernel bandwidth
size_t hmlp::gofmm::CommandLineHelper::d |
(Optional)
double hmlp::gofmm::CommandLineHelper::h = 1.0 |
(Optional) set the default Gaussian kernel bandwidth.
DistanceMetric hmlp::gofmm::CommandLineHelper::metric = ANGLE_DISTANCE |
(Default) geometric-oblivious scheme.
size_t hmlp::gofmm::CommandLineHelper::n |
end CommentLineSupport() Basic GOFMM parameters.
double hmlp::gofmm::CommandLineHelper::stol = 1E-3 |
(Default) user-defined approximation toleratnce and budget.