https://github.com/siddk/memory

Core repository for all research code centered around using Memory Networks for question-answering.
https://github.com/siddk/memory

Last synced: 3 months ago
JSON representation

Core repository for all research code centered around using Memory Networks for question-answering.

• Host: GitHub
• URL: https://github.com/siddk/memory
• Owner: siddk
• Created: 2016-04-09T21:17:43.000Z (about 9 years ago)
• Default Branch: master
• Last Pushed: 2016-04-22T17:02:31.000Z (about 9 years ago)
• Last Synced: 2025-01-12T10:32:24.314Z (4 months ago)
• Language: Python
• Homepage:
• Size: 4.48 MB
• Stars: 3
• Watchers: 3
• Forks: 2
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Memory

Core repository for all research code centered around [Memory Networks](http://arxiv.org/pdf/1410.3916v11.pdf)

for question-answering. This directory is broken up in the following way:

1. BABI Tasks (/babi)

    - Replicate Question-Answering Code from [End-to-End Memory Networks](http://arxiv.org/pdf/1503.08895v5.pdf)

      on Facebook AI's corpus of [20 BABI Tasks](https://research.facebook.com/researchers/1543934539189348)

2. BABI LSTM Model (/naive_rnn)

    - Experiment to see if a Recurrent Neural Network (LSTM) Model with question + story sentence fed

      together can match original BABI Memory Network Results. Sums and averages all word embeddings

      for queries and stories, rather than keeping all the data together.

3. BABI Sequential LSTM Model (/sequential_rnn)

    - Experiment to see if feeding the complete information (flattened story sentence + query 

      embeddings) into an RNN can give good results on the first three bAbI Tasks. In this case,

      each timestep fed into the RNN will be a combination of a sentence of the story, and the

      query.