Skip to content

Latest commit

 

History

History
26 lines (21 loc) · 797 Bytes

TODO.md

File metadata and controls

26 lines (21 loc) · 797 Bytes
Raw

NN Architecture

  1. Experiment: add layers
  2. Implement and Experiment: Change non-linearities

Features

  1. Implement: Remove stop-words, weight word vectors by TF-IDF
  2. Experiment: Dimension of word vectors
  3. *** Implement and Experiment: Headline/Body Cosine Similarity Feature (append to preds in add_pred_op) ***

Training word vectors

  1. Experiment: do not train word vectors
  2. Experiment: train word vectors after N epochs, N a hyperparameter

Optimization

  1. Experiment: tune batch size
  2. Implement and Experiment: tune learning rate
  3. Implement and Experiment: change optimizer
  4. Implement and Experiment: add regularization
  5. Experiment: add dropout

Evaluation

  1. Print training error (after N epochs)

Debugging

  1. Verify that scoring in model.py is correct