DLS5 Week 2 Assignment 2 Exercise 2, Step 1.5 - Examining Test Set Performance - clarification needed

Course Q&A Deep Learning Specialization Sequence Models
1

Hi experts,
I am not sure if I understand what Step 1.5 should really produce as far as grading in concerned.

From the description, it suggests that the word “cherish” should get a heart emoji, right?

My training result shows:

Epoch: 0 --- cost = 410.4336578831472
Accuracy: 0.5454545454545454
Epoch: 100 --- cost = 63.61263974696143
Accuracy: 0.9318181818181818
Epoch: 200 --- cost = 0.7391301193275178
Accuracy: 1.0
Epoch: 300 --- cost = 0.31048254133339553
Accuracy: 1.0

and test set performance as follows:

Training set:
Accuracy: 1.0
Test set:
Accuracy: 0.9107142857142857

However, the test sentence “i cherish you” returns me a “smiley” emoji, not a “heart”.

I am not sure if the grading process will specifically test for similar use case and interpret it as a
fail. Appreciate if the mentors could clarify a bit, and obviously right now my sentiment is a
“disappointed” emoji seeing the wrong emoji return for the word “cherish” in the sentence.

Rgds,
MCW

3

Thanks for quick response Tmosh and clearing my doubt which I thought might be something
wrong with my codes. All good now have gone ahead and submitted it!!

Cheers,
MCW

4

The description never suggests that the word “cherish” should get a heart emoji. It said:

  • You can check that the word “cherish” does not appear in the training set.
  • Nonetheless, let’s see what happens if you write “I cherish you.”
    i cherish you :smile:
    i love you :heart:
    funny lol :smile:
    lets play with a ball :baseball:
    food is ready :fork_and_knife:
    not feeling happy :disappointed:

Amazing!

  • Because cherish has a similar embedding as love, the algorithm has generalized correctly even to a word it has never seen before.
5

Ah ok makes sense looks like i’ve misinterpreted it. Thanks.