DLS Course 5 Week 2 Assignment 1 Debiasing: Neutralize

Course Q&A Deep Learning Specialization Sequence Models
1

Hi everyone,

I have somehow unexpected results in the additional part of the exercise: “neutralize(word, g, word_to_vec_map)”

After debiasing I get following result: “cosine similarity between receptionist and g, after neutralizing: -1.521716345503026e-17”
Basically it’s not not as expected -4.4e-17, although it is small.

The formula says: build the dotproduct of e,g and multiply it with g, devided by the squared norm of g. The norm is the same as in previous code, which I expect correct - based on the autograder.

Hopefully, since it’s ungraded, I can provide code? @Mentors

Please help, i’m irritated.

2

I expect this to be a rounding error,… if I multiply with g after dividing by the squared norm the result is: “-3.8581292784004314e-17”.

Can one cofirm this is correct or at least important?

3

Can this be related to: