CODEWITHSUNDEEP
pythonnumpymatrixlinear-algebrasympy
Flying pig
Flying pig

Reputation: 612

How to convert the 'dtype = object' NumPy array into 'dtype = float' array?

This problem is caused by using SymPy to find reduced row echelon matrix, I have screen capture below. By using SymPy, I can get a NumPy array of object, it is quite confusing for me here. First, I don't see numbers in the rref matrix is wrapped by ' ' or " "，how come they are 'object'. Second, is there any way to convert it into a 'dtype = float' np.array?

M = sy.Matrix([[4, 0, 11, 3], [7, 23, -3, 7], [12, 11, 3, -4]]) 
M_rref = M.rref()
M_rref = np.array(M_rref[0])
M_rref.astype(float)

enter image description here

Upvotes: 0

Views: 2097

Answers (2)

gust
gust

Reputation: 945

Short answer:

M_rref.astype(float) will fix your problem.

Long answer:

Why does this happen?

In order to find out, try going back to your matrix M and try the following:

>>> print(M[0])
>>> print(type(M[0]))

You will get:

4
sympy.core.numbers.Integer

That's why. Even though it says it's 4, it's actually a SymPy object that is stored inside the SymPy matrices. And that's the way it should be: SymPy is for symbolic mathematics. Therefore, when you convert it to a numpy array, NumPy will recognize sympy.core.numbers.Integer as an object.

Hope that answers your question.

Upvotes: 1

smichr
smichr

Reputation: 18939

In this PR there is a note about incorrect conversion of numpy matrices to SymPy matrices. Using that (and not float, which I don' think is intended to convert contents of an object to floats) I would do:

>>> sy.Matrix(M.rows, M.cols, [sympify(x) for x in M_rref]).n()

This should give you a SymPy matrix with entries evaluated as floats.

Upvotes: 0

Related Questions