Coder Social home page Coder Social logo

neuralese's Introduction

Translating Neuralese

This paper contains code and experiments for our ACL 2017 paper:

J Andreas, A Dragan and D Klein. Translating Neuralese. ACL 2017. https://arxiv.org/abs/1704.06960

Getting the data

Driving

The data we collected for the driving game can be found in data/drive. The game server itself is located in src/server.

Colors

Data comes from

B McMahan and M Stone. A Bayesian Model of Grounded Color Semantics. TACL 2017.

cd data/color
wget http://paul.rutgers.edu/~bcm84/rugstk_v1.0.tar.gz
tar xzf rugstk_v1.0.tar.gz 
mv rugstk_v1/data/munroecorpus/* .
rm -r rugstk_v1

Birds

Data comes from:

P Welinder, S Branson, T Mita, C Wah, F Schroff, S Belongie and P Perona. Caltech-UCSD Birds 200. Technical Report CNS-TR-2010-001, California Institute of Technology.

S Reed, Z Akata, H Lee, and B Schiele. Learning deep representations of fine-grained visual descriptions. CVPR 2016.

(I am trying to figure out where the file of annotations I have came from and whether it is public. In the meantime, contact me directly at [email protected] if you want to run bird experiments.)

Running the experiments

In the root of the project, first:

mkdir experiments

To run an experiment, do:

./run.sh <config>

where <config> is the configuration file for the experiment in question (located in the configs directory). Logs will be written to the appropriate experiment dir. In general you probably want to run a train_X script followed by an eval_X script.

Evaluation

This will appear in the log file when evaluation is turned on in the config. The belief-matching criterion from our paper is called rkl, while the translation criterion is called dot. Lines of the form [c,l:model] show results of the pragmatic evaluation: the first number is reward and the second is task completion. Lines of the form [c-l] show results of the semantic evaluation. c and l designate neuralese and natural language speakers respectively.

Visualization

You probably also want to see what the system is saying! Evaluation runs will create a file called vis.json. If you place this file in the same directory as vis.html (located in the project root), you can look at the computed beliefs associated with both natural language and neuralese messages.

neuralese's People

Contributors

jacobandreas avatar

Watchers

paper2code - bot avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.