Status: In Maintenance.
Code for Artificial neural networks for neuroscientists: A Primer
This repository provides tutorial-style code for training artificial neural networks on simple neuroscience-relevant tasks, and for analyzing these networks using a variety of neuroscience methods.
(1) Jupyter notebook for training convolutional neural networks, representation-similarity-analysis (RSA), and gradient-based tuning analysis.
(2) Jupyter notebook for training LSTMs on a memory task, and visualizing neural activity.
(3) Jupyter notebook for training RNNs on a working memory task, and fixed-point-based dynamical system analysis.
(4) Jupyter notebook for training Excitatory-Inhibitory RNNs on a decision-making task, and analyzing network connectivity.
This code is tested with Python 3 and a recent version of Pytorch. Please kindly let me know if it does not work for you. All code can be run quickly (a few minutes) on a laptop CPU.
Please email me at gyyang.neuro@gmail.com or send issues/pull requests, if you have any feedback. Thank you!