deeplearning-sequences

Experiments on RNN, LSTM etc Please have a look at the following word2vector playgrounds:

1. Word Embedding Visual Inspector. See here
   
2. word2vector demo See here

File: rossmann_timeSeries_noExtData.ipynb
Objectives:

          i) Feature engineering on Time Series data
             Elapsed-event-time and Rolling summaries
         ii)Categorical embeddings
        iii) Using fastai on tabular data
         iv) Understanding 1-cycle policy

File: 2_simple_rnn_IMDB.ipynb
Objective(s):

    A. Familiarising with Document processing
       using gensim.
    B. Convert tokens with each document to corresponding
       'token-ids' or integer-tokens.
       (For text cleaning, pl refer wikiclustering file
       in folder: 10.nlp_workshop/text_clustering)
       (Keras also has  Tokenizer class that can also be
       used for integer-tokenization. See file:
       8.rnn/3.keras_tokenizer_class.py
       nltk can also tokenize. See file:
       10.nlp_workshop/word2vec/nlp_workshop_word2vec.py)
    C. Creating a Bag-of-words model
    D. Discovering document similarity

File: tf data API
Objectives:

       i)  Learning to work with tensors<br>
       ii) Learning to work with tf.data API<br>
      iii) Text Classification--Work in progess<br>

File: textClassification_bidirectional_LSTM.ipynb

Objectives:
i) Learning to work with tfds.load
ii) Learning to work with tf.data API
iii) Text Classification--WORK IN PROGRESS
(but works perfectly)

File: 0_basic_document_processing.ipynb
Objective(s):

    A. Familiarising with Document processing
       using gensim.
   B. Convert tokens with each document to corresponding
      'token-ids' or integer-tokens.
      (For text cleaning, pl refer wikiclustering file
      in folder: 10.nlp_workshop/text_clustering)
      (Keras also has  Tokenizer class that can also be
      used for integer-tokenization. See file:
      8.rnn/3.keras_tokenizer_class.py
      nltk can also tokenize. See file:
      10.nlp_workshop/word2vec/nlp_workshop_word2vec.py)
   C. Creating a Bag-of-words model
   D. Discovering document similarity

File: kingMinusWoman.ipynb
Objective(s)

Experimentation with pre-created word2vec file
Works with gensim 3.8.3
Test:

  paris− france+germany (should be close to Berlin)
  bought - bring + seek (should be close to sought)