Thesis/ADNC
1
0
Fork 0
mirror of https://github.com/JoergFranke/ADNC.git synced 2024-11-17 13:58:03 +08:00
Code Issues Packages Projects Releases Wiki Activity

Update README.md

Browse Source
This commit is contained in:
Jörg Franke 2018-07-11 18:08:53 +02:00 committed by GitHub
parent 868cd77743
commit 01eef3a890
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 5 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
README.md
View File

@ -6,8 +6,8 @@
This repository contains a implementation of a Differentiable Neural Computer (DNC) with advancements for a more robust and This repository contains a implementation of a Differentiable Neural Computer (DNC) with advancements for a more robust and
scalable usage in Question Answering. It is published on the MRQA workshop at the ACL 2018. In this repository the ADNC is applied to the scalable usage in Question Answering. It is published on the MRQA workshop at the ACL 2018. This advanced DNC (ADNC) is applied to the
[20 bAbI QA tasks](https://research.fb.com/downloads/babi/) with [state-of-the-art results](#babi-results) adn the [20 bAbI QA tasks](https://research.fb.com/downloads/babi/) with [state-of-the-art results](#babi-results) and the
[CNN Reading Comprehension Task](https://github.com/danqi/rc-cnn-dailymail) with [CNN Reading Comprehension Task](https://github.com/danqi/rc-cnn-dailymail) with
[passable results](#cnn-results) without any adaptation or hyper-parameter tuning. [passable results](#cnn-results) without any adaptation or hyper-parameter tuning.
@ -21,12 +21,6 @@ The repository contains the following features:
- The following advancements to the DNC: - The following advancements to the DNC:
<!--
This repository is the groundwork for the MRQA 2018
paper submission [Robust and Scalable Differentiable Neural Computer for Question Answering](https://arxiv.org/abs/1807.02658). It contains a modular and
fully configurable DNC with the following advancements: -->
<table> <table>
<tbody> <tbody>
<tr> <tr>
@ -70,10 +64,10 @@ fully configurable DNC with the following advancements: -->
</tbody> </tbody>
</table> </table>
Please find more information about the advancements and the experiemnts in Please find more information about the advancements and the experiments in
- MRQA 2018 paper submission [Robust and Scalable Differentiable Neural Computer for Question Answering](https://arxiv.org/abs/1807.02658) - MRQA 2018 paper submission [Robust and Scalable Differentiable Neural Computer for Question Answering](https://arxiv.org/abs/1807.02658)
- Master thesis about the [Advanced DNC for Question Answering](http://isl.anthropomatik.kit.edu/cmu-kit/downloads/Master_Franke_2018.pdf) - Master thesis about the [Advanced DNC for Question Answering](http://isl.anthropomatik.kit.edu/cmu-kit/downloads/Master_Franke_2018.pdf) with a detailed DNC description.
The plot below shows the impact of the different advancements in the word error rate with the bAbI task 1. The plot below shows the impact of the different advancements in the word error rate with the bAbI task 1.
@ -173,4 +167,4 @@ Possible models are `dnc`, `adnc`, `biadnc` on bAbi Task 1 and `biadnc-all`, `bi
| Stanford AR | 72.2 | 72.4 | | Stanford AR | 72.2 | 72.4 |
| AoA Reader | 73.1 | 74.4 | | AoA Reader | 73.1 | 74.4 |
| ReasoNet | 72.9 | 74.7 | | ReasoNet | 72.9 | 74.7 |
| GA Reader | 77.9 | 77.9 | | GA Reader | 77.9 | 77.9 |