Course 5 Week 2 Assignment 1 Exercise 2

Course Q&A Deep Learning Specialization Sequence Models
1

I am having a hard time to understand how to compute the cosine similarity in Excercise 2 - complete analogy

From the template # Compute cosine similarity between the vector (e_b - e_a) and the vector ((w’s vector representation) - e_c)

How to best compute the second argument "(w’s vector representation - e_c) when calling the cosine similarity subroutine?
I used: ‘word_to_vec_map[w] - e_c’ but get an error when I run the complete_analogy_test

However, I am not sure if the error is related to this line of code?

TypeError Traceback (most recent call last)
in
33 print("\033[92mAll tests passed")
34
—> 35 complete_analogy_test(complete_analogy)

in complete_analogy_test(target)
26 word_to_vec_map[key] = np.array(word_to_vec_map[key])
27
—> 28 assert(target(‘a’, ‘a_nw’, ‘c’, word_to_vec_map) == ‘c_nw’)
29 assert(target(‘a’, ‘a_s’, ‘c’, word_to_vec_map) == ‘c_s’)
30 assert(target(‘a’, ‘synonym_of_a’, ‘c’, word_to_vec_map) != ‘c’), “Best word cannot be input query”

in complete_analogy(word_a, word_b, word_c, word_to_vec_map)
44 # If the cosine_sim is more than the max_cosine_sim seen so far,
45 # then: set the new max_cosine_sim to the current cosine_sim and the best_word to the current word (≈3 lines)
—> 46 if None > None:
47 max_cosine_sim = None
48 best_word = None

TypeError: ‘>’ not supported between instances of ‘NoneType’ and ‘NoneType’

2

That is the correct code for the 2nd parameter that is sent to consine_similarity(…)

1 Like
3

The problem seems to be that you haven’t completed the last three lines in the complete_analogy() function (where all those “None” statements are).

4

hi TMosh, i did cosine_sim = cosine_similarity(e_b-e_a, word_to_vec_map[‘c_w’]-e_c) and complete the function as you said, but still got error , could you have a check?

TypeError Traceback (most recent call last)
in
33 print("\033[92mAll tests passed")
34
—> 35 complete_analogy_test(complete_analogy)

in complete_analogy_test(target)
26 word_to_vec_map[key] = np.array(word_to_vec_map[key])
27
—> 28 assert(target(‘a’, ‘a_nw’, ‘c’, word_to_vec_map) == ‘c_nw’)
29 assert(target(‘a’, ‘a_s’, ‘c’, word_to_vec_map) == ‘c_s’)
30 assert(target(‘a’, ‘synonym_of_a’, ‘c’, word_to_vec_map) != ‘c’), “Best word cannot be input query”

in complete_analogy(word_a, word_b, word_c, word_to_vec_map)
41 print(word_to_vec_map[w])
42 print(np.shape(word_to_vec_map[w]))
—> 43 cosine_sim = cosine_similarity(e_b-e_a, word_to_vec_map[w]-e_c)
44
45 # If the cosine_sim is more than the max_cosine_sim seen so far,

TypeError: unsupported operand type(s) for -: ‘str’ and ‘str’

5

Your case seems to be simple, since you tried to subtract “string” from “string”. The most possible error is you did not convert “string” to “vector” by using “word_to_vec_map”.

6

Most likely the issue is in the code you implemented for this comment instruction:
# Get the word embeddings e_a, e_b and e_c (≈1-3 lines)