Code for Active sensing in the categorization of visual patterns

Note

• These code may be incomplete, as they are extracted from a bigger directory.
• Only the first layer of directory structure is kept (for now).
• Only matlab/c functions are kept (no data, fig, etc. for now).
• The folder correlation-shuffle contains the latest correlation analysis.
• The folder exp-setup contains some of the earliest code for setting up the experiment.

The Bayesian active sensing core

• a_Get_BALDscoreProp.m computes the BAS score map. Among other functions, it calls a_Model.m and a_GetBALDGM_mex.c.
• a_Model.m computes type posteriors, gradients of parameters, and other stuff.
• a_GetBALDGM_mex.c computes the BAS score of individual location.

Making Figure 3A in the paper (testing)

• Run BAS/analysis final/simu_BASs.m.
  • Comment out lines 65-66. Uncomment lines 47, 49-50.
  • No input needed. Output is SIM.
• Run BAS/analysis final/revmap_sim.m.
  • Comment out lines 138-139, 142-172, 212. Uncomment lines 107-135, 210.
  • DRemovePhase(D) is introduced during the review process.
  • Input is SIM. Outputs are [REV, RrevMap, RdrevMap].
• Run BAS/analysis final/plot_rev_maps.m.
  • Inputs are RrevMap, RdrevMap. Output is Figure 3A.