50+ days of Machine Learning

 Hello :) Today is Day 57! Quick summary of today: Got introduced to word vectors with lecture 1  + small showcase Read the paper and took notes about: Efficient Estimation of Word Representations in Vector Space (word2vec) Distributed Representations of Words and Phrases and their Compositionality (negative sampling) First I watched the lecture by Professor Manning, and then I read the papers so some of the material was overlapping, but there were still interesting new parts in each of the three.  1) Lecture notes My notes from the lecture were not that long so I will first share them. After the lecture, there was a small colab document I could run through.  In the lecture, we mention about matrix multiplication and the larger the dot product between the 2 words, the more similar they are -> exp(u0T @ vc) So in the colab there was a model loaded. And I searched for the word embeddings of bread and baguette which in my view are similar words.  In the below pic, if a 2 values from bo

 Hello :) Today is Day 56! A quick summary of today: Looked at  Natural Language Processing with Transformers Found Stanford Uni's cs224n NLP with DL  course Today... was a bit weird, I suppose. My goal was to start something related to NLP and build on my knowledge from the previous days.  Before that, I will share about the book (from the 1st point). I saw this book about NLP with transformers, and looking at the content it looked awesome, covers a variety of topics in NLP - transformers, text classification, named entity recognition, text generation, summarization, question answering, dealing with different data.  So I decided to give it a go, and took notes on the 1st chapter, which was great. (you can see them at the end of this post)  I learned about how the transformer came to be, its 'story'. And my notes are below.  Then the book starts to use huggingface - a popular library that provides a vast amount of models, datasets, tokenizers and NLP tools. But I feel like