---

SVDTOOL.CC

---

Source:

svdtool.cc

Target:

svdtool.o

Type:

library

Version: 0.1 Date: 8/15/2000

Los Alamos in-house version

Version: 0.9b Date: 2/9/2001

Beta production release

Make the subroutines more robust and introduce XML style documentation for the source code.

Packages:

svdman ( documentation )

Parts:
svdman.cc ( documentation )
svdtool.cc ( documentation )
svd.h ( documentation )
libclasvd.a ( documentation )
libF77.a ( documentation )

Includes:

svd.h

Blocks:

WriteRawData()
Kind:
method
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov

Comments:
Write some matrix data to an output file
Usage:
Inputs:
Outputs:
Returns:
Parameters:
string ofname; // Name of output file

vector<double> dat; // Data matrix

int numcols; // Number of columns

vector<string> row_labels; // Row labels

vector<string> col_labels; // Column labels


splint()
Kind:
method
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov
Numerical Recipes in C

Comments:
Cubic spline interpolation. Modified from Numerical Recipes in C.
Usage:
Inputs:
Outputs:
Returns:
Parameters:
double* xa; // Pointer to ordinate values for interpolation

double* ya; // Pointer to abscissa values for interpolation

double* y2a; // Pointer to spline coefficients

int n; // Number of data points in xa[],ya[],y2a[]

double x; // X (ordinate) value for interpolation

double& y; // Reference to interpolated value


spline()
Kind:
method
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov
Numerical Recipes in C

Comments:
Cubic spline coefficients calculation. Modified from Numerical Recipes in C
Usage:
Inputs:
Outputs:
Returns:
Parameters:
double* x; // Pointer to ordinate values

double* y; // Pointer to abscissa values

int n; // Number of elements in x[] and y[]

double yp1; // 1st derivative constraint at low x. If greater than 1.e30, it s a natural spline with 2d derivative = 0

double ypn; // 1st derivative constraint at high x. If greater than 1.e30, it s a natural spline with 2d derivative = 0

double* y2; // Pointer to array of output spline coefficients


Max()
Kind:
method
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov

Comments:
Return the greater of two integers
Usage:
Inputs:
Outputs:
Returns:
Type: int
The greater of the input values
Parameters:
int m; // Value #1

int n; // Value #2


Min()
Kind:
method
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov

Comments:
Return the lesser of two integers
Usage:
Inputs:
Outputs:
Returns:
Type: int
The lesser of the input values
Parameters:
int m; // Value #1

int n; // Value #2


SplitLine()
Kind:
method
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov

Comments:
Split a string at the positions of an input character, ala Perl split(). Return the number of break points found.
Usage:
Inputs:
Outputs:
Returns:
Type: int
The number of break points (i.e. characters) found in the string.
Parameters:
string strline; // The input string to be fragmented.

int start; // The start position for the search. The first position = 0.

int nwords; // The maximum number of words to extract.

char ch; // The character that defines the break points.

vector<string>& words; // The output array of words from the split.


Data::Data(string)
Kind:
constructor
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov
Patricia Dyck, Los Alamos National Laboratory/tdyck@lanl.gov

Comments:
Constructs an instance of the data class, initializing the matrix using data from the input file named ifname.
Usage:
Inputs:
Outputs:
Returns:
Parameters:
string ifname; // The file name to be read for initializing Data()


Data::Data(int, Data&)
Kind:
constructor
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov
Patricia Dyck, Los Alamos National Laboratory/tdyck@lanl.gov

Comments:
Construct an instance of the class Data() based on an existing class. The new instance has one column missing from the data matrix, defined by an input index.
Usage:
Inputs:
Outputs:
Returns:
Parameters:
int index; // The column number to be deleted from the source matrix.

Data& d; // The source class to be read in creating the new instance.


Data::Data(Data&)
Kind:
constructor
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov
Patricia Dyck, Los Alamos National Laboratory/tdyck@lanl.gov

Comments:
Copy constructor for Data() Copies the matrix from the source.
Usage:
Inputs:
Outputs:
Returns:
Parameters:
Data& d; //


Data::NumCols()
Kind:
method
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov

Comments:
Return the number of columns in the data matrix
Usage:
Inputs:
Outputs:
Returns:
Parameters:

Data::NumRows()
Kind:
method
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov

Comments:
Return the number of rows in the data matrix
Usage:
Inputs:
Outputs:
Returns:
Parameters:

Data::WriteCorr(string, vector<double>)
Kind:
method
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov

Comments:
Output the correlations from the SVD evaluation to a file.
Usage:
Inputs:
Outputs:
Returns:
Parameters:
string out_base; // Output filename base

vector<double> correlations; // Vector array of calculated correlation values.


Data::WriteGroups(string, double)
Kind:
method
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov

Comments:
Determine the SVD groups and write them to an output file.
Usage:
Inputs:
Outputs:
Returns:
Parameters:
string out_base; // Output filename.

double thresh; // Scale for the threshold value for inclusion within a group.


Data::SubMeanRow()
Kind:
method
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov

Comments:
Subtract the mean row value from each row of the data matrix.
Usage:
Inputs:
Outputs:
Returns:
Parameters:

Data::SVD()
Kind:
method
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov
Patricia Dyck, Los Alamos National Laboratory/tdyck@lanl.gov

Comments:
Calculate the SVD of the data matrix, placing results in U, VT, S and B. Also generate labels for the SVD modes/
Usage:
Inputs:
Outputs:
Returns:
Parameters:

Data::MinusSVD()
Kind:
method
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov

Comments:
Invert the signs of U, VT and B.
Usage:
Inputs:
Outputs:
Returns:
Parameters:

Data::JacknifeData(double, vector<double>&)
Kind:
method
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov
Patricia Dyck, Los Alamos National Laboratory/tdyck@lanl.gov

Comments:
Perform evaluation on the current data column. This method is called within the reduced matrix class.
Usage:
Inputs:
Outputs:
Returns:
Parameters:
double xval; // The ordinate value for interpolation of the missing column.

vector<double> data_col; // The original column of data.


Data::EvaluateSVD(vector<double>&)
Kind:
method
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov
Patricia Dyck, Los Alamos National Laboratory/tdyck@lanl.gov

Comments:
Perform evaluation of the SVD using the jacknife method.
Usage:
Inputs:
Outputs:
Returns:
Parameters:
vector<double> correlations; // Output list of Pearson correlation values from the evaluation.


Data::CalcColData(vector<double>&,vector<double>&)
Kind:
method
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov
Patricia Dyck, Los Alamos National Laboratory/tdyck@lanl.gov

Comments:
Calculate the "missing" data column by matrix multiplication.
Usage:
Inputs:
Outputs:
Returns:
Parameters:
vector<double>& calcValues; // The calculated interpolated column from the SVD.

vector<double>& data_col; // The output reconstructed data column.


InterpolateSVD(double,vector<double>&)
Kind:
method
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov
Patricia Dyck, Los Alamos National Laboratory/tdyck@lanl.gov

Comments:
Interpolation of the missing column of the SVD.
Usage:
Inputs:
Outputs:
Returns:
Parameters:
double xval; // The ordinate value for interpolation (missing assay label).

vector<double>& newValues; // Vector of values for the elements of the interpolated column.


Data::CorrColData(int, vector<double>&,vector<double>&)
Kind:
method
Authors:
Michael Wall, Los Alamos National Laboratory/mewall@lanl.gov
Patricia Dyck, Los Alamos National Laboratory/tdyck@lanl.gov

Comments:
Calculate the Pearson correlation between the reconstructed column and the original data column
Usage:
Inputs:
Outputs:
Returns:
Parameters:
int index; // Column number in the original data matrix.

vector<double>& data_col; // Reconstructed data column

vector<double>& correlations; // List of calculated correlations to which the current is added

Build:

Dependencies:
svd.h

Compile:
g++ -O2 -I./F2CLIBS -c svdtool.cc
Platforms:
RedHatLinux version 7.0

Compilers:
gcc version 2.96

RedHatLinux version 6.1

Compilers:
egcs version 1.1.2