Skip to content

TensorFlow implementation of "Tracking the World State with Recurrent Entity Networks".

License

Notifications You must be signed in to change notification settings

jimfleming/recurrent-entity-networks

Repository files navigation

Recurrent Entity Networks

This repository contains an independent TensorFlow implementation of recurrent entity networks from Tracking the World State with Recurrent Entity Networks. This paper introduces the first method to solve all of the bAbI tasks using 10k training examples. The author's original Torch implementation is now available here.

Diagram of recurrent entity network architecture

Results

Percent error for each task, comparing those in the paper to the implementation contained in this repository.

Task EntNet (paper) EntNet (repo)
1: 1 supporting fact 0 0
2: 2 supporting facts 0.1 3.0
3: 3 supporting facts 4.1 ?
4: 2 argument relations 0 0
5: 3 argument relations 0.3 ?
6: yes/no questions 0.2 0
7: counting 0 0
8: lists/sets 0.5 0
9: simple negation 0.1 0
10: indefinite knowledge 0.6 0
11: basic coreference 0.3 0
12: conjunction 0 0
13: compound coreference 1.3 0
14: time reasoning 0 0
15: basic deduction 0 0
16: basic induction 0.2 0
17: positional reasoning 0.5 1.7
18: size reasoning 0.3 1.5
19: path finding 2.3 0
20: agents motivation 0 0
Failed Tasks 0 ?
Mean Error 0.5 ?

NOTE: Some of these tasks (16 and 19, in particular) required a change in learning rate schedule to reliably converge.

Setup

  1. Download the datasets by running download_babi.sh or from The bAbI Project.
  2. Run prep_data.py which will convert the datasets into TFRecords.
  3. Run python -m entity_networks.main to begin training on QA1.

Major Dependencies

  • TensorFlow v1.1.0

(For additional dependencies see requirements.txt)

Thanks!

  • Thanks to Mikael Henaff for providing details about their paper over Thanksgiving break. :)
  • Thanks to Andy Zhang (@zhangandyx) for helping me troubleshoot numerical instabilities.
  • Thanks to Mike Young (@mikalyoung) for providing results on some of the longer tasks.

Releases

No releases published

Packages

No packages published