Tuesday, July 3, 2012

Feedback 03-07-2012

Progress
  • Corrected and improved the implementation of the formulas for the univariate and bivariate normal distribution
    • Univariate: 




    • Bivariate:



  • Effects:
    • Simulations are more stable
    • Simulations are much faster
  • Added visualization of the belief space for grid worlds



Results (Preview)
  • Experimental Set-up:
    • Light Dark domain with the following layout:
********L*
******A*L*
********L*
********L*
********L*
***G****L*
********L*
********L*
********L*
********L* 
  • Initial belief: uniform over all possible states, except the goal state
  • # Episodes: 5000
  • Optimal is: two steps right, four steps down, five steps left (probably not the true optimal line)











No comments:

Post a Comment