DeepNLP.md

DeepNLP

0. Intro

   • backprop,
   • theano,
   • mnist classification;

1. Embeddings

   • Text classification: bag of words, TF-IDF
   • NLTK: lemmatization, stemming;
   • w2vec:skip gram & cbow;
   • fasttext, glove;

2. Convolutions

   • Text classification with embeddings;
   • Avito classification task;
   • (need to update)

3. RNN

   • RNN,
   • backprop through time,
   • gradient explosion and vanishing,
   • LSTM & GRU,
   • gradient clipping

4. Seq2Seq

   • Seq2Seq architecture
   • Attention mechanisms
   • Image Captioning
   • Modern architectures overview
   • (need to update)