search

scijs / contributing

Contribution guide lines
10 stars 0 forks source link

Singular value decomposition & eigendecompositions #8

Open mikolalysenko opened 9 years ago

mikolalysenko commented 9 years ago

Dense SVD is probably the most asked for feature in ndarray, and it would be nice to have some module which robustly and efficiently computes them for arbitrary data. Standard applications include data analysis, machine learning and statistics.

Related to this, it would be nice to have an eigendecomposition modules (for things like solving polynomial roots and so on), though it is less critical.

The priority here would be numerical robustness more than efficiency, though faster is of course always nice.

Also it would be good to have some options for iterative eigensolvers, like Arnoldi/Lanczos iteration.

rreusser commented 9 years ago

I'll work toward this, but I expect numerical robustness to take some work. I recall there being subtlety and room for choices when using shifts. Should just be a matter of extending and refining though.