CODEWITHSUNDEEP
pythonnumpy
rudy
rudy

Reputation: 196

parameters constraint in numpy lstsq

I'm fitting a set of data with numpy.lstsq():

numpy.linalg.lstsq(a,b)[0]

returns something like:

array([ -0.02179386,  0.08898451,  -0.17298247,  0.89314904])

Note the fitting solution is a mix of positive and negative float.

Unfortunately, in my physical model, the fitting solutions represent a mass: consequently I'd like to force lstsq() to return a set of positive values as a solution of the fitting. Is it possible to do this?

i.e.

solution = {a_1, ... a_i, ... a_N} with a_i > 0 for i = {1, ..., N}

Upvotes: 9

Views: 4259

Answers (1)

Alicia Garcia-Raboso
Alicia Garcia-Raboso

Reputation: 13913

Non-negative least squares is implemented in scipy.optimize.nnls.

from scipy.optimize import nnls

solution = nnls(a, b)[0]

Upvotes: 18

Related Questions