Hi, guys!
I got stuck in the assignment for Week 2.
The Naive Bayes model I constructed could output 0.0 and 9165
for print(logprior) and print(len(loglikelihood)), respectively.
However, testing of my model failed as shown in the picture above.
I think my implementation for the variable, freq_pos could have some problems.
Could you please give me some advice?
It’s hard to advice from just this error. I would comment that it is indicative that train_naive_bayes() function returns wrong loglikelihood value (even though the length of it is correct).
From the error it is visible that loglikelihood['sunglass'] has a key error (no sunglass in your loglikelihood dictionary). Maybe when you calculated the log likelihood of the word at the very end you forgot to assign it to loglikelihood[word]? Maybe in the for loop you did not loop through vocab? It could easily been something else.
If you are still stuck and cannot find a solution, you could private message me your Assignment notebook attached. (how to download your notebook)
For anyone in the future: If you encounter the same problem, pay close attention to
# Increment the number of positive words by the count for this (word, label) pair
N_pos += ?
# else, the label is negative
else:
# increment the number of negative words by the count for this (word,label) pair
N_neg += ?