Copyright (C) 2023 Marjorie Xie and Ashok Litwin-Kumar

This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.

This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with this program.  If not, see <http://www.gnu.org/licenses/>.

SUMMARY:
This repository contains code used in:
M. Xie, S.P. Muscinelli, K.D. Harris, & A. Litwin-Kumar. Task-dependent optimal representations for cerebellar learning. eLife (2023).

The components are
(1) script_simulation.py: this script walks through how to set up our model of the granule cell layer presynaptic to a Purkinje cell, how to generate Gaussian process targets, and how to train the model to learn a target. 
At the top of the script are a list of parameters with default values which you can play with.

(2) script_theory.py: this script walks through the analytic calculation of generalization error given a Gaussian process target and a kernel. These scripts call functions from the libraries below:

(3) networks.py: functions for generating activity in the granule cell layer and fitting the readout weights of the network, given a target. 

(4) sphlib.py: functions needed for performing the decomposition of a kernel function and targets in the spherical harmonic basis.

(5) kernellib.py: functions for computing the kernel of ReLU network analytically and for computing generalization error analytically. 

(6) targets.py: functions for generating Gaussian process targets. 

This software uses Python (python.org). Tested on Python 3.6. 

CONTACT:
mx2183@columbia.edu
ak3625@columbia.edu