NLP
-
Understanding the gradient flow in a computation graph by building a simple feedforward network
-
Stepping through the training of a neural network while building a sentiment classification model
-
Word representation using the common bag of words model
-
The predecessors of Transformers: RNNs, how they work, and why attention was the missing block
-
Problems with pure RNNs and GRUs, and how LSTM tries to solve them
-
Building intuition for attention by understanding the limitations of RNNs
-
Positional Encoding: one of the most clever engineering decisions in the Attention Is All You Need paper