Plaut_Model

Dependencies

See requirements.txt

General Information

There are several directories required for proper functionality:

1. checkpoints This directory stores any created checkpoints for future use.
2. code This directory contains the code and also the configuration file.
3. dataset This directory contains the Plaut, anchor, and probe datasets in .csv format.
4. results This directory stores the simulation results folders, which contain the plots and .csv.gz files created with each simulation.

How to Use

1. The configuration file for simulations is located in code/config.cfg. This file contains all user-settable parameters, with descriptions of the usage of each parameter. This file must not be renamed nor moved in order for the simulation to run successfully. With each simulation, a copy of the config.cfg file will be placed in the simulation results folder.
2. After all desired parameter settings are made in the configuration file, the code in the Jupyter Notebook file plaut_model.ipynb can be executed. Information will be printed and plotted on intervals specified in the configuration file.
3. A folder containing simulation results will be created inside the results directory. This simulation folder will be named with the format <label>-S<seed>D<dilution>O<order>-<date>. Inside the simulation folder will be loss and accuracy plots, a copy of the configuration file, as well as .csv.gz file containing accuracy information for each word at specified intervals during the simulation.