Adaptive Q Learning Citizen Science-Sophie
#notesFromPaper Year : Tags : q learning cartoscope cairns #cairnsRL Authors: Spatharioti
Q_Learning_Citizen_Science-Sophie had Q learning, but it didn’t adapt to player skill on the fly
generate data by simulating players from a variety of player models
0.95 discount factor, 0.1 learning rate. Used weighted Q picking
The state was the levels that had just been played and the number of mistakes that had been made in those levels (to decrease state space, players were categorized as G,B,U instead of having it be a limitless space of numbers)
