xinsui77 / cvar_jmathpsych_2021

Overview

This repository contains the data and code needed to reproduce the main results from our paper:

C.Gagne and P.Dayan, Peril, prudence and planning as risk, avoidance and worry. Journal of Mathematical Psychology (2021) 102617, https://doi.org/10.1016/j.jmp.2021.102617.

Installations

The code has only a few requirements, which can installed by running:

conda env create -f environment.yml

Code organization

The code is organized according to the sections of the paper.

Code for:

• calculating CVaR for a single choice (Figure 1) is in 'cvar_single_choice'.
• plotting the nCVaR/pCVaR tree (Figure 2) is in 'cvar_tree'
• evaluating nCVaR/pCVaR for a random policy (Figure 3) is in 'policy_evaluation'
• calculating the optimal nCVaR/pCVaR policies (Figures 4-5) is in 'policy_optimization'
• obtaining the optimal greedy replay sequences (Figures 6-7) are in 'replay optimization'

Policy evaluation, optimization, and replay shares code in the folder 'shared'. Here, you can find the nCVaR/pCVaR Bellman evaluation and optimality operators.

Simulation results

All of the simulations have already been run and the results are stored in 'simulation_results'. The '.mat' files were created using code privately borrowed from Chow et al. 2015 (not contained in the repo); these files contain some information about the MDPs, which is reused by our scripts and also were used to check our implementation versus theirs. These should not be deleted. All of the other '.npy' files can be recreated.

If you have any questions, please feel free to contact [e-mail].