Week 3. Peaks and Troughs.

Mostly troughs. I started this week endeavoring to write my own bag of words model; having studied several models last week. However, on Wednesday, I realized that not only did I not know *how* to write my own train method, but my mentor and I had different ideas about how to implement the model. I was using a Naive Bayes Classifier, whereas he thought an implementation using a neural net might be better. This was a bit of a confidence killer as I thought I was ready. I spent the rest of the week doing the FastlAi course on embeddings, as well as searching other courses on bag of word models using neural nets. I also read more papers on attention mechanisms applied to parse trees.

Next week, I’m hoping to progress towards reading literature specific to transformers, as well as drawing high-level representations of different mini-problems. Before the week is over, however, I’d like to finish my bag of words model alongside a pictorial representation of it.