h2r / pomdp-py

A framework to build and solve POMDP problems. Documentation: https://h2r.github.io/pomdp-py/
MIT License
210 stars 49 forks source link

Fix action selection when visitation count is 0 #7

Closed jasonxyliu closed 3 years ago

jasonxyliu commented 4 years ago

in POMCP and POUCT, initialize num_visits to 0 instead of 1 in _ucb, add check num_visits == 0, Q-value set to inf

zkytony commented 4 years ago

Thanks! But could you only commit the fix to the po_uct.pyx and pomcp.pyx files? Also there's a commented out print statement that you probably forgot to remove.

zkytony commented 3 years ago

This is fixed. See 6f06c9ebec5.