Modular Symbolic (and indexed) Matrix Creation with Sympy - python-2.7

Question

I am trying to create a symbolic and modular python function using sympy that takes in the dimensions j (rows) & k (columns) and returns a matrix indexed. Here is the function:
import sympy
from sympy import *
from sympy import init_printing,Matrix
from sympy.physics.vector import dynamicsymbols
import numpy as np
init_printing()
def weights_func_L(j,k):
#
# jth nuerons in the k ( or L - 1 ) layer
# j rows
# k columns
#in other words, this is the weight (in a specific layer j) for the jth nueron and the kth neuron in the
#previous (L-1 layer) i.e wjk
#this will return a weight matrix for j rows (i.e corresponds to the # of layers)
#and k (corresponds to the # of nuerons)
#
matrix_bin = ones(j,k)
matrix_bin = flatten(matrix_bin)
basic_w_string = 'w'
for a in np.arange(len(matrix_bin)):
#symbols(basic_w_string+str(countj)+str(countk))
if j == k:
matrix_bin[a] = symbols(basic_w_string+str(a // k)+str( a % j))
else:
matrix_bin[a] = symbols(basic_w_string+str(a // k)+str( a % j))
#print matrix_bin[a]
# for aa in np.arange(j):
# matrix_bin[countj,countk] = symbols(basic_w_string+str(countj)+str(countk))
# countj += 1
# countk += 1
matrix_bin2 = np.reshape(matrix_bin,(j,k))
return Matrix(matrix_bin2)
This works for a square matrix:
weights_func_L(3,3)
which yields:
However for a non square matrix such that j!=k:
weights_func_L(5,3) yields
similarly calling weights_func_L(3,5) yields:
any guidance would be much appreciated!

Answer 1

The a % j should have been a % k, like this:
def weights_func_L(j,k):
matrix_bin = ones(j,k)
matrix_bin = flatten(matrix_bin)
basic_w_string = 'w'
for a in np.arange(len(matrix_bin)):
matrix_bin[a] = symbols(basic_w_string+str(a // k)+str(a % k))
matrix_bin2 = np.reshape(matrix_bin,(j,k))
return Matrix(matrix_bin2)
>>> weights_func_L(5, 3)
⎡w₀₀ w₀₁ w₀₂⎤
⎢ ⎥
⎢w₁₀ w₁₁ w₁₂⎥
⎢ ⎥
⎢w₂₀ w₂₁ w₂₂⎥
⎢ ⎥
⎢w₃₀ w₃₁ w₃₂⎥
⎢ ⎥
⎣w₄₀ w₄₁ w₄₂⎦