using numpy's cumprod functionality with matrix multiplication

Asked May 08 '20 at 10:53

Active May 08 '20 at 11:46

Viewed 242 times

I'm trying to use the functionality of numpy's cumprod, but for matrices. Having a matrix filled with homogeneous transformations (here random for simplicity) I want to accumulate the transformations using numpy's matrix multiplication.

n = 1000
data = np.random([n, 4,4])
dataAcc = cumProd(data, np.matmul)

dataAcc should contain the following:

dataAcc[0,:,:] = data[0,:,:]
dataAcc[1,:,:] = data[0,:,:] @ data[1,:,:] = dataAcc[0,:,:] @ data[1,:,:]
dataAcc[2,:,:] = dataAcc[1,:,:] @ data[2,:,:]
...

Is there a way to do this using fast numpy functions or some equivalent? I do not want to use loops...

Cheers

edited May 08 '20 at 11:46

asked May 08 '20 at 10:53

wodtko

Duplicate with no answer: https://stackoverflow.com/questions/59492474/cumulative-dot-product-with-numpy – Warren Weckesser May 08 '20 at 11:01
Another unanswered dup: https://stackoverflow.com/questions/61512885/cumulative-result-of-matrix-multiplications – Warren Weckesser May 08 '20 at 11:07
One more, with an answer, but the discussion appears to focus on having just four matrices: https://stackoverflow.com/questions/27993153/how-to-fold-accumulate-a-numpy-matrix-product-dot – Warren Weckesser May 08 '20 at 11:13
I added an answer over in https://stackoverflow.com/questions/59492474/cumulative-dot-product-with-numpy – Warren Weckesser May 08 '20 at 11:47

using numpy's cumprod functionality with matrix multiplication

0 Answers0