Skip to content
master
Go to file
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
 
 
 
 
 
 

README.md

MGpi: A Computational Model of Multiagent Group Perception and Interaction

Synthetic Dataset (Group Layouts)

https://www.dropbox.com/s/e1rzcbeeu86eh96/interaction_data.zip?dl=0

Dependency Installations

This installs dependencies in a virtual environment, inheriting system packages. Run chmod u+x setup.sh; ./setup.sh

Simulating Communication Interactions

This will generate training data for our MGpi Network

  1. Download and extract the Synthetic Dataset linked above
  2. Run source virt_mgpi/bin/activate. This activates the virtual environment with dependencies installed.
  3. Run python simlulate.py. This imports GroupInteract from sim_interact.py or sim_interact_dyn.py, and simulates group communications.

Training policy

  1. After you have simulations, chane appropriate variables in train.py and run nohup python train.py simulations_directory model_name number_of_epochs.
  2. model_name here can be:

AllHist for MGpi

All-NoSelfState for baseline NSO

NoNeighMod for baseline SSO

All-NoKPMGate for baseline EQPOOL

All-SocPool for baseline SOCPOOL

Testing policy

After changing appropriate variables in test.py, run nohup python test.py

Testing performance on group detection

After changing appropriate variables in test_group.py, run nohup python test_group.py

About

MGpi: A Computational Model of Multiagent Group Perception and Interaction

Resources

Releases

No releases published

Packages

No packages published

Languages