Sum for Poisson Equation #498
Description
Description:
This equation indicates that we should iterate through every word in vocab (which we define as every unique word in the training set). First off, this is unclear, and it is incorrect. Following this direction leads to incorrect probabilities that we are supposed to match in Problem 5.
Suggested Fix:
I found that if you iterate through each unique word in each message, this does return the correct probabilities. In addition, it aligns with the statement in the problem to use a default value if you encounter a word that is not in the training set.
Volume & Lab (By Name)
Volume 3: NaiveBayes
Source Lines
Relevant segments in `/Volume3/NaiveBayes/NaiveBayes.md`:
- Approximate line: 274:
- Section: Training the Model:
- TeX: $\argmax_{k\in K}\ln\left(P(C=k)\right) +
\sum_{i\in\text{Vocab}}
\ln\left(\frac{(r_{i,k}n)^{n_i}e^{-r_{i,k}n}}{n_i!}\right),$ (`\argmax_{k\in K}\ln\left(P(C=k)\right) +
\sum_{i\in\text{Vocab}}
\ln\left(\frac{(r_{i,k}n)^{n_i}e^{-r_{i,k}n}}{n_i!}\right),`)