Implementation of the deep neural networks described in Kuperwajs, Schütt, and Ma (2023).
This repository implements networks as models of human play in 4-in-a-row using PyTorch. The networks are trained and tested on human decisions in large-scale data. Trained networks are available upon request, and the cognitive model code is available at https://github.com/ionatankuperwajs/4IAR-improvements.
network.py
: architecture for the neural networkspreprocessing.py
andcustom_dataset.py
: data formattingload_train.py
andtraining.py
: training scriptsload_test.py
andtesting.py
: testing scriptsanalysis.py
andsummary_stats.py
: analysis scripts for figures in the papermodel_preprocessing.py
,model_comparison.py
, andmodel_improvements.py
: data formatting and analysis for the cognitive model comparisontrain_network.sh
,train_network_array.sh
, andtest_network.sh
: bash scripts for training and testing on the computing cluster