invert singular matrix on python

Question

I have to read multiple data from csv files, and when I want to invert matrix from csv data, I get this:

numpy.linalg.linalg.LinAlgError: singular matrix

and the process get stuck on this section :

J = np.mat([dtdx,dtdy,dtdz]).transpose()
dd = np.mat(ttcal-tt)
dm = (scipy.linalg.inv(J.transpose()*J))*((J.transpose())*(dd.transpose()))

and data from "J" like this :

[[-6.81477651e-03 -7.90320450e-03  6.50533437e-05]
 [-6.71080644e-03 -6.00135428e-03  6.50533437e-05]]

and data from "dd" like this :

[[0.00621772 0.00537531]]

i has check this data and i find this :

tes = J.transpose()*J

and the result like this :

[[ 9.14761019e-05  9.41324993e-05 -8.79884397e-07]
 [ 9.41324993e-05  9.84768945e-05 -9.04538042e-07]
 [-8.79884397e-07 -9.04538042e-07  8.46387506e-09]]

I need to invert this matrix but this data is singular matrix. I have to try on matlab r2017b and running well.

I need to solve this problem on python.

isn't the definition of a Singular Matrix, a square matrix that does not have a matrix inverse? — Reblochon Masque
– Reblochon Masque, Commented Sep 12, 2018 at 7:17
@ReblochonMasque Yes it is. There is something called Psuedo Inverse for such matrices. I think he is referring to that — Sreeram TP
– Sreeram TP, Commented Sep 12, 2018 at 7:19
@WahyuHadinoto If you want a psuedo inverse use linalg.pinv from numpy — Sreeram TP
– Sreeram TP, Commented Sep 12, 2018 at 7:21
@WahyuHadinoto linalg.pinv works for the matrix you posted as you can see from my answer. There may be other cases in your problem where the generated matrices are such that psuedo inverse can'be determined. Also, make sure that your numpy and scipy are latest versions.. — Sreeram TP
– Sreeram TP, Commented Sep 12, 2018 at 7:40

Eypros · Accepted Answer · 2018-09-12 07:22:28Z

5

Have you tried using pseudo-inverse numpy.linalg.pinv instead? It's supposed to deal with these situations.

B = np.linalg.pinv(a)

But I would suggest to check that you really calculated correctly your matrix and a singular matrix is supposed to appear.

answered Sep 12, 2018 at 7:22

Eypros

5,7437 gold badges54 silver badges89 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

Sreeram TP · Accepted Answer · 2018-09-12 07:25:35Z

2

If you are sure the calculations you made are correct and it is what you wanted, then you can go for a psuedo inverse for the singular matrix that you are having.

It can be done in python like this..

mat = np.array([[ 9.14761019e-05,  9.41324993e-05, -8.79884397e-07],
 [ 9.41324993e-05,  9.84768945e-05, -9.04538042e-07],
 [-8.79884397e-07, -9.04538042e-07,  8.46387506e-09]])

p_inv = np.linalg.pinv(mat)

print(p_inv)

# output

array([[-1.00783988e+13,  5.50963966e+11, -9.88844703e+14],
       [ 5.50963966e+11, -3.01194390e+10,  5.40580308e+13],
       [-9.88844703e+14,  5.40580308e+13, -9.70207468e+16]])

answered Sep 12, 2018 at 7:25

Sreeram TP

12k8 gold badges65 silver badges121 bronze badges

Collectives™ on Stack Overflow

invert singular matrix on python

2 Answers 2

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related