This is the code for the ICLR 2023 Submission, 'Actionable Neural Representations: Grid Cells from Minimal Constraints'.

Included in the code are:

1. 6 files that perform optimisation in finite periodic of very large periodic spaces of 1, 2, or 3 dimensions. These should just run and output saved lists of the relevant parameters which can be used for plotting

   1. Line.py
   2. Circle.py
   3. Torus.py
   4. Plane.py
   5. Volume.py
   6. PeriodicVolume.py

2. A folder which acts as a module, NRT_functions (NRT = Neural Representation Theory), which contains three files

   1. helper_functions.py - just some helpful functions
   2. losses.py - the losses used
   3. plotter.py - some helpful plotting functions

3. A jupyter notebook that lets you plot the output of a simulation - Plotter.ipynb

The simulation runs are automatically saved in a file system organised by data ("YYMMDD") and time ("HHMMSS"). The plotter uses these to reload the data and plot the learnt representations

Required packages:

1. Jax
2. Numpy
3. matplotlib