Hello I have just recently started working with python 3d arrays and I am not too familiar with how to multiply each block in the 3d array with all other blocks in the 3d array. I am working with a np array with shape (31, 22, 16). Let say my numpy array is x with shape (31, 22, 16) I will like to get x[16:31,0,0] and multiply it with every other x[16:31,i,j]. For each pair of multiplication I will like to sum the result like so np.sum(x[16:31,i,j] * x[16:31,i,j]).
The inefficient code uses two nested loops it becomes really slow when I have large data. Could someone please help me with a faster solution for computing the below code please? To be more exact I want to use np.einsum to do this but I am not sure how it works. Thanks!
llList = []
for gg in range(nrGates):
for bb in range(nrBeams):
llList.append((bb, gg))
# posInx = [16->31] contains numbers from 16 to 31
for ll in range(nCells): # nCells = 352
llAI = llList[ll] # contains tuples of (nrBeams, nrGates)
spectL = currentData.spectrum[posInx, llAI[0], llAI[1]] # shape (16,)
# currentData.spectrum has shape (31, 22, 16)
for mm in range(nCells):
mmAI = llList[mm]
spectM = currentData.spectrum[posInx, mmAI[0], mmAI[1]]
currentData.Dlm[ll, mm] = np.sum(spectL * np.conj(spectM))
