DL : Basic Concept of RNN
Part 5.1 of Deep Learning Specialization
Published in
3 min readApr 5, 2020
1. Why RNN, not NN ?
1.1 RNN cell
1.2 Cell vs Layer
1 Cell, 1 Layer
Sequential input into the same RNN cell.
N Cell, N-Layer >> Deep RNN
2. Bidirectional RNN
3. RNN — GRU — LSTM
3.1 RNN
3.2 LSTM
4. Word Representation
- One-hot Encoding
- Word Embedding
- Word2Vec
4.1 One-hot Encoding
represent words by Sparse Matrix of all unique words
4.2 Word Embedding
represent words by matrix of selected features
4.3 Word2Vec
learn word embedding using Neural Network
(Encoder — Decoder)
5. Attention Model
learn where (when) to focus according to input