Assignment week2 exercise5 propagate

Course Q&A Deep Learning Specialization Neural Networks and Deep Learning
1

Hello everyone,

In exercise 5 - propagate, I am getting the following output (including some logging output of mine between BEGINNING OF FUNCTION BODY and END OF FUNCTION BODY) including an assertion error about the values of dw:

BEGINNING OF FUNCTION BODY
m: 3
Shape of w: (2, 1)
Shape of X: (2, 3)
Shape of Y: (1, 3)
Shape of Z: (1, 3)
Shape of A: (1, 3)
Shape tmp_cost_val: (1, 3)
Cost: 5.873473260859346
END OF FUNCTION BODY
dw = [[0.93406047]
 [2.18899374]]
db = 0.06593791286065272
cost = 5.873473260859346
BEGINNING OF FUNCTION BODY
m: 3
Shape of w: (2, 1)
Shape of X: (2, 3)
Shape of Y: (1, 3)
Shape of Z: (1, 3)
Shape of A: (1, 3)
Shape tmp_cost_val: (1, 3)
Cost: 5.873473260859346
END OF FUNCTION BODY
---------------------------------------------------------------------------
AssertionError                            Traceback (most recent call last)
<ipython-input-91-be03942b69a5> in <module>
     14 print ("cost = " + str(cost))
     15 
---> 16 propagate_test(propagate)

~/work/release/W2A2/public_tests.py in propagate_test(target)
     37     assert type(grads['dw']) == np.ndarray, f"Wrong type for grads['dw']. {type(grads['dw'])} != np.ndarray"
     38     assert grads['dw'].shape == w.shape, f"Wrong shape for grads['dw']. {grads['dw'].shape} != {w.shape}"
---> 39     assert np.allclose(grads['dw'], expected_dw), f"Wrong values for grads['dw']. {grads['dw']} != {expected_dw}"
     40     assert np.allclose(grads['db'], expected_db), f"Wrong values for grads['db']. {grads['db']} != {expected_db}"
     41     assert np.allclose(cost, expected_cost), f"Wrong values for cost. {cost} != {expected_cost}"

AssertionError: Wrong values for grads['dw']. [[0.93406047]
 [2.18899374]] != [[0.99845601]
 [2.39507239]]

while the expected output should be:

Expected output

dw = [[0.99845601]
 [2.39507239]]
db = 0.001455578136784208
cost = 5.801545319394553

So first, there seems to be a mistake with my computation of cost and then, there seems to be an issue with my gradients dw and db. I have tried several different ways of computing cost and dw which I would like to show here, but I am not sure if I would violate the coursera honor code then. How should we proceed here?

Thanks a lot for your help.

2

You’re right that we’re not supposed to share source code publicly. But we can do one more step in debugging (at least). If you look at the formulas for dw and db, they aren’t that complicated. But they depend on the values of A and m, so it’s worth also checking those. I added some prints like yours to show those values and here’s what I get:

m = 3
A = [[0.99987661 0.99999386 0.00449627]]
dw = [[0.99845601]
 [2.39507239]]
db = 0.001455578136784208
cost = 5.801545319394553
m = 3
A = [[0.99987661 0.99999386 0.00449627]]
All tests passed!

What do you see when you print those values? I’ll bet that your A value is wrong. Note that is also an input to the cost formula which is also slightly off.

3

Dear @paulinpaloalto,

Thanks a lot for your advice. Indeed you were right that my value of A was wrong. My mistake was that I added the bias term b to X instead of adding it to to the result of $w^T X$. Python broadcasting is great, but in this case, it caused me a lot of headaches. I definitely learnt to pay close attention to these details subtle in the future :slight_smile:

1 Like
4

Cool! Glad to hear that you found the solution. Yes, I’ve had that same thought about broadcasting before as well: it’s great when it saves you work, but it’s also dangerous in that something that you didn’t intend can “just work” instead of throwing a “size mismatch” error. Oh, well! As you say, there is no substitute for attention to detail. :nerd_face: