operands could not be broadcast together with shapes (2,2) (2,100000)?

Question

Trying to convert a matlab code to python and get the following error when trying to calculate the first and second data:

operands could not be broadcast together with shapes (2,2) (2,100000) 

when trying to rewrite this matlab snippet in vsc:

data=u*(s.^(1/2))*(randn(2,Ndata)); data=data+Ey;

---

The code that I have converted to python is:

Ky11 = 10
Ky22 = 1
rho = 0
Ky12 = np.sqrt(Ky11 * Ky22) * rho
Ky = np.array([[Ky11, Ky12], [Ky12, Ky22]])

Ey1 = 0
Ey2 = 0
Ey = np.array([[Ey1], [Ey2]])

Ndata = 100000

u, s, v = np.linalg.svd(Ky)
# Create a 2x2 diagonal matrix with the singular values of Ky (s) on the diagonal
s_matrix = np.diag(s)

# Compute the square root of each element of s_matrix and then multiply
# with u and np.random.randn(2, Ndata)
# Transpose u so that the dimensions match for element-wise multiplication
data = u.T * np.sqrt(s_matrix) * (np.random.randn(2, Ndata))
data = data + Ey

Ky_estimate = np.cov(data.T)

can anyone help with a solution?

I tried to compute the square root of s_matrix element-wise and then multiply with transposed u for dimesnions reasons but didn't get what I expected which was to get the code running without any errors. Have searched other stack overflow pages with "same" question but didn't solve my problem.

Answer 1

Are you trying to do matrix multiply? Use @ instead of * . Big Gotcha in Python. The operator * is element by element multiplication.