2 ms·
Word embeddings are also "simply computed." If you use GloVe, then the vectors are obtained by factoring a matrix of co-occurence counts. The difference betwee
by yorwba 8y ago
Word embeddings are also "simply computed." If you use GloVe, then the vectors are obtained by factoring a matrix of co-occurence counts.
The difference between machine learning and "simple" feature extraction is mostly just in the choice of metaphors used to describe the computation, not in any fundamental properties.
- ma2rten 8y agoRight. Naive Bayes is considered to be a machine learning algorithm, but also consists of just "simple counting".