RNNs are specialized architectures for sequential data. The key idea is to leverage parameter sharing which has two benefits : the sequence length can be one not found in the training set and that we maintain the statistical power of learned features across all time steps (we don’t have to learn the same features again and again).
Linear Least Squares
Notes from the book Thinking Funtionally in Haskell
This article is mostly adapted from here. However, the article is laboriously long and not enough meat. This is, what I hope, a condensed version.
I used the following tutorials to create this site:
subscribe via RSS