Skip to content

Code for comment on Tajima et al. (2019). Optimal policy for multi-alternative decisions. Nature Neuroscience.

License

Notifications You must be signed in to change notification settings

DiODeProject/MultiAlternativeDecisions

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

55 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

MultiAlternativeDecisions

Code for comment on Tajima et al. (2019). Optimal policy for multi-alternative decisions. Nature Neuroscience.

These codes modify the analysis presented in the paper by Tajima et al. (2019), to examine the implication of maximising geometric-discounted future rewards on the optimal policy, rather than maximising arithmetic mean reward rate.

N.B. reward rate across trials is not maximised, but set to 0; this has the effect of considering only single trial dynamics, since the cost of waiting for the next trial becomes zero.

Installation

All scripts are written in MATLAB and have been tested with MATLAB R2018b. They should also run on earlier versions of MATLAB.

To run the scripts, download them, extract them into a folder of your choice, and navigate to this folder within MATLAB.

Usage

To use any of the script in the fig folders, you need to first point MATLAB to the shared utility functions by calling

addpath('shared/')
addpath('shared/bads-1.0.4/')

at the MATLAB command line. See the individual fig folders for further usage instructions.

To switch between maxisation of geometrically-discounted rewards and arithmetic mean reward rate, toggle the value of the geometric variable.

To scale option values change the value of the valscale variable.

To change between linear and various nonlinear utility functions uncomment the appropriate utilityFunc definition.

About

Code for comment on Tajima et al. (2019). Optimal policy for multi-alternative decisions. Nature Neuroscience.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • MATLAB 73.5%
  • Fortran 18.7%
  • HTML 3.9%
  • C++ 2.8%
  • Mathematica 0.7%
  • Makefile 0.2%
  • Other 0.2%