Right now in linalg.py, I'm using numpy.linalg.inv() for matrix inverses.
I'd like to implement it from scratch instead, but it looks rather difficult...
Note that Cholesky decomposition is only applicable to Hermitian matrices, which will work for PCA but hopefully I can also implement a more general-case algorithm like LUP decomposition as a fallback
Right now in
linalg.py
, I'm usingnumpy.linalg.inv()
for matrix inverses. I'd like to implement it from scratch instead, but it looks rather difficult...Cholesky decomposition looks tractable, see: