Why numpy function inner and numpy function dot return different results ?

Asked Jan 28 '16 at 14:37

Active Apr 16 '16 at 10:58

Viewed 49 times

I do not understand why the numpy functions inner and dot do not return the same results on these inputs :

Z1 = [ 0.20802286  0.23876957  0.21183842  0.23651123  0.22368603  0.24748229
  0.21947122  0.19641609  0.23148395  0.20815766  0.20839586  0.23032438
  0.22005879  0.22142259  0.21964123  0.2236328   0.24725002  0.22615054
  0.22185984  0.2242185 ]

Z2 = [ 0.10577075  0.33866838 -0.12644889 -0.16111459  0.02895356  0.05370375
  0.18242676 -0.06122486 -0.21797925 -0.06018776 -0.36965305  0.31710415
  0.18896685  0.08465404  0.47444968 -0.07607215 -0.21495446  0.0677735
 -0.36636134 -0.21047364]

sum(Z1[:]*Z2[:]) = 9.7144514654701197e-17

dot(Z1[:]*Z2[:]) = 9.7144514654701197e-17

inner(Z1[:]*Z2[:]) = 6.9388939039072284e-17

edited Apr 16 '16 at 10:58

asked Jan 28 '16 at 14:37

augustin-barillec

2

Python doesn't have `dot` or `inner` functions. Are you using Numpy? – PM 2Ring Jan 28 '16 at 14:41
2

This is a floating point precision issue All floating point operations have limited precision and errors do accumulate. – Alex Jan 28 '16 at 14:41
Yes, these are numpy functions. – augustin-barillec Jan 29 '16 at 15:42
what i do not understand is that the definition of inner(a,b) is precisely sum(a[:],b[:]) and yet the results are different on the inputs Z1 and Z2 above. – augustin-barillec Jan 29 '16 at 15:43

Why numpy function inner and numpy function dot return different results ?

0 Answers0