Matmul between a matrix and a vector using CKKSVector

[mokachocolat]

Question

Hello, what is the equivalent of

A = np.zeros((256,1))
B = np.zeros((2,1))
A @ B.T

when A is a CKKSVector and B also ?

Further Information

I'm trying to code a backward function in which I need to compute the derivative of a weight matrix of size (256,2) by multiplying my encrypted input x (vector with shape (256,)) and a variable delta (vector with shape (2,)).

I tried to use .transpose(), .matmul() and reshape delta after a call of .decrypt() but it does not work for me.

It seems like, .matmul() works only when the first shape of both matrix are the same but it seems like we do not have all tools to do so with CKKSVectors.

System Information

• OS: Ubuntu
• OS Version: 22.04
• Language Version: Python 3.10.12