python - Numpy array, indexing and convolving confusion -

i'm trying complete following function, have been running problems indexing, resulting in "valueerror: operands not broadcast shapes (0,9) (5)".

i think error might coming how i'm trying call values ssd_difference[], i'm not entirely sure.

also how go using convolve2d based on hint given below? understand numpy has function it, have no idea need put in make work.

additional information: binomialfilter5() returns 5x1 numpy array of dtype float representing binomial filter. i'm assuming "weights[]" ssd_difference[] values.

def transitiondifference(ssd_difference): 

""" compute transition costs between frames, taking dynamics account.

instructions: 1. iterate through rows , columns of ssd difference, ignoring first 2 values , last 2 values.
1a. each value @ i, j, multiply binomial filter of length 5 (implemented later in code) weights starting 2 frames before until 2 frames after, , take sum of products.

            i.e. weights frame are:                  [weight[i - 2, j - 2],                   weight[i - 1, j - 1],                   weight[i, j],                   weight[i + 1, j + 1],                   weight[i + 2, j + 2]] 

multiply binomial filter weights @ each i, j output.

it may take little bit of understanding why computing this, simple explanation change frame 4 5, lets call ch(4, 5), , make weight:

ch(4, 5) = ch(2, 3) + ch(3, 4) + ch(4, 5) + ch(5, 6) + ch(6, 7)

this accounts weights in previous changes , future changes when considering current frame.

of course, weigh these sums binomial filter, weight ch(4, 5) still important one, gives better understanding.

args:     ssd_difference (numpy.ndarray): difference matrix produced                                     ssd function.  returns:     output (numpy.ndarray): difference matrix takes preceding ,                             following frames account. output                             difference matrix should have same dtype                             input, 4 rows , columns smaller,                             corresponding frames have valid                             dynamics.  hint: there efficient way 2d convolution. think       coordinates using consider preceding ,       following frame pairings. """  output = np.zeros((ssd_difference.shape[0] - 4,                    ssd_difference.shape[1] - 4), dtype=ssd_difference.dtype) # write code here. in range(len(ssd_difference)):     j in range(len(ssd_difference)):         if == 0:             if j > 1:                 output[i,j] = np.sum( ssd_difference[i-2:i+2]*binomialfilter5())         elif == ssd_difference.shape[0] - 1:             if j < ssd_difference.shape[1] - 2:                 output[i,j] = np.sum( ssd_difference[i-2:i+2]*binomialfilter5())         else:             output[i,j] = np.sum( ssd_difference[i-2:i+2]*binomialfilter5()) # end of function. return output 

as commented, should tell line produced error message.

but can guess, since there couple of lines operation involves broadcasting. is:

output[i,j] = np.sum( ssd_difference[i-2:i+2]*binomialfilter5()) 

you write binomialfilter5() produces (5,1) array, error talks (5,). doesn't matter here, should keep number of dimensions straight. (5,1) signficantly different (5,).

output has shape (ssd_difference.shape[0] - 4, ssd_difference.shape[1] - 4). iterating i,j both on range(len(ssd_difference)). output[i,j] result in index error. when iterating on 2d array, better use correct shape element, rather len().

but suspect immediate error results ssd_difference[i-2:i+2]. when i==0, ssd_difference[-2:2]. producing (0,9) array, since -2 index means second last, larger 2.

i think intending pull 5 rows array, match 5 values in other array. correct iteration, think be:

for in range(output.shape[0]):     j in range(output.shape[1]):        ....        output[i,j] = np.sum(ssd_difference[i:i+5, :] * binomialfilter5())        ... 

you should test expressions individually in interactive shell, selected values of i. ssd_difference[i:i+5, :] should have shape (5,9), , binomialfilter5() should (5,1).