Linear Regression

efficiently linear regression O( n^2 log( √n/ε ) ) Time algorithm ( ε-accuracy ).
an Hebrew analyze of the time running could be found here : https://github.com/dudupo/linear-regression/blob/master/svd.pdf

I have ‘copy-pasted’ the document as it was, directly from assignment, so it might seem strange (for who is’t an exercises examiner). I hope to update the manner in future.

currently, it’s not working (yet), and has a buggy nature. As a first step I focusing on demonstrate a polynomial regression. after that I planning to improve the code by constant Time factor, and ‘righter cpp literary’.

Issuses

(I) reordering the vectors of the orthonormal matrix

the svd decompaction, calculating the vectors by ascending order ( matched to the eigenvalue ), currently there is no regard to the subject while assembling back.

(II) results of polynomial regression aren’t even similar to the expectations.

Testing

note that you might want to decrease the size of the matrix.

mingw32-make
python ./scripts/svdtest_gen.py
python ./scripts/svdtest_run.py