Principal component analysis software, and linear algebra library

v0.9

Start from initial linear algebra stub (qr.c)

v1.0

Fully working PCA version. All the linear algebra (matrix) functions are integrated All the data are stored statically, in the stack. the mlatrices are stored as two dimensional arrays, therefore requiring fixed size allocation. So each matrix or vector takes up MAX_COLxMAX_COL size in stack memory Data sets are limited to 150 linesa running the programm typically requires 2 Gb of memory!

v1.2

All data sets (in main and in functions) are allocated dynamically and stored in the heap. Also using syslinGauss instead of syslinQR uses much less memory (clearly visible with valgrind). However, size of matrices and data set is still fixed (inevitable because matrices are two-dimensional arrays). This versions runs v1.2 with 350 rows x 350 cols maximum. Above this, valgrind detects invalid memory reads and writes. Also corrected a bug in str2headers (added a space at the end of linestr)

v2.0

Complete overhaul Programme split between a linear algebra library (libal.c) and a main pca.c principal componet analysis programme. libal.c groups basic statistics functions (mean, standard deviation, sumsquares,..), matrix structure definition (including dimensions n,m and pointer to one-dimensional array, to be allocated dynamically), and all the matrix operations including transpose, scaling, porduct, inversion, linear system resolution, QR transformation, eigenvalues, eigevctors, etc..

pca.c is the main PCA resolution programme. It reads data from a default input programme "pca_input.dat", calculates the pca and displays it screen and text file.

v2.1 -- Uses libstring library function 'tokenize()' instead of internal 'str2headers()' to read headers. v2.1.2 is the first fully working version (without memory leaks) that uses libstring.c.