Model from:
Tsuda B, Richmond BJ, Sejnowski TJ. Exploring strategy differences between humans and monkeys with recurrent neural networks. https://doi.org/10.1371/journal.pcbi.1011618
rnn_model.py contains recurrent neural network model trained by reinforcement learning.
humanVmonkey_env.py contains the specifications for the three working memory tasks from Wittig et al. 2016 (http://learnmem.cshlp.org/content/23/11/644).
Organization of rnn_model.py is
- helper fxns
- definition of network class
- definition of worker class
- train fxn
- get_experience fxn
- test fxn
- main
- definition of parameters and output directories
- creation of central network
- creation of worker objects that run the network
- script to deploy workers for training AND testing
Command to run rnn_model.py:
python3 rnn_model.py [NETSZ] [EPS_TO_TRAIN_ON] [PERFTHRESH] [RUNNO] [GPU]