ardaegeunlu / x-armed-bandits Goto Github PK

Hi, we have built a new library for different X-armed bandit algorithms, https://github.com/WilliamLwj/PyXAB. Could you try it and see whether our implementation of HOO is satisfying?

Thank you!

Hi, thanks for the update. I tried running the code after just replacing the file, and now I get the following problem.

KeyError Traceback (most recent call last)
in ()
22 x_armed_bandit.set_time_horizon(max_plays=1000)
23 x_armed_bandit.set_environment(environment_function=testFunction.draw_value)
---> 24 x_armed_bandit.run_hoo()
25
26 # this is the most rewarding point explored so far after "max_plays" rounds.

~\Documents\X-armed-Bandits-master\HOO.py in run_hoo(self)
149 # now we have selected the most promising child from the tree we had,
150 # we can draw an arm in X from the coverings
--> 151 reward = self.selection_of_arm_from_covering(current_node_depth, current_node_in_level)
152 self.create_children_of_node(current_node)
153

~\Documents\X-armed-Bandits-master\HOO.py in selection_of_arm_from_covering(self, h, i)
71 :return: received reward.
72 """
---> 73 subset_space = self.tree_coverings(h, i)
74 # now we need to arbitrarily select one arm from this subset of the space.
75 # the chosen strategy is selecting the midpoint of the subset but this can

~\Documents\X-armed-Bandits-master\Partitioner.py in halve_one_by_one(self, height, place_in_level)
65 parent_h = height - 1
66 parent_i = (place_in_level + 1) / 2
---> 67 parent_space = self.covering_sequence[parent_h, parent_i]
68 #print("index = {0},{1} and parent = {2},{3}".format(height,place_in_level,parent_h,parent_i))
69 # partition the dimension_to_be_partitioned into half.

KeyError: (0, 1.5)

Hi, tried to run your code, but it seems like Partitioner.py is exactly the same as HOO.py. Is this intended?