Congratulations!

Recurrent Neural Networks (RNNs) for Language Modeling with Keras

David Cecchini

Data Scientist

Wrap-up

• Introduction to language tasks:
  • Sentiment classification
  • Multi-class classification
  • Text Generation
  • Neural Machine Translation
• Sequence to sequence models
• Implementation in Keras

RNN pitfalls and different cell types

• Vanishing and exploding gradient problems
• GRU and LSTM cells
• Word vectors and the Embedding layer
• Better sentiment analysis

Multi-class classification

• Data preparation
• Transfer learning
• Keras models
• Model performance

Text generation and NMT

• Text Generation
  • Chars as token
  • Data preparation
  • Generate sentences mimicking Sheldon
• Neural Machine Translation
  • Words as tokens
  • Data preparation: encoders and decoders
  • Translate Portuguese to English

Congratulations!!!

Recurrent Neural Networks (RNNs) for Language Modeling with Keras