Code for the paper "Improving with Practice: A Neural Model of Mathematical Development" and the eponymous paper in TopiCS. Presentation slides for the CogSci2016 talk can be found here. The Latex source code for the paper can be viewed on Overleaf.
Python is required to run these scripts. The scripts were written in Python 2, but should work in Python 3.
The Python packages required can be installed with pip install -r requirements.txt
.
The specific version of the Nengo Python package was ceaf387.
with_feedback.py
is the final full-network versioncounting_only.py
uses only the counting networkbetter_pred_run.py
generates the data from the prediction in the paperhetmem_learning.py
andautoens_learning.py
generate the data for the comparison of learning with and without Voja
All analysis was performed in Jupyter notebooks, which end with the file extension ipynb
.
These can be opened by running jupyter notebook BLAH.ipynb
where BLAH.ipynb
is the name of the notebook you
desire to view. However, at this time, all the files required to duplicate the analysis seen in the
paper are not currently available, due to some bad backing up on my part. I'm currently working on retrieving
the files from my old desktop.
- Clean up the data files and acquire missing files
- Clean up the creation of the data files and logs
- Make applicable scripts compatible with the Nengo version of PyTry, instead of commenting out chunks of code whenever they are run in Nengo GUI.