def step(self, action, goal):
'''
Action to discover a node.
If action is -1, we stop discovery and return reward as influence maximization output else reward is 0
'''
if action == -1:
self.done = True
self.reward_ = parallel_influence(self.graph,
full_graph=self.fullGraph,
influence=self.influence_algo,
times=self.times_mean)
if self.normalize:
self.reward = self.reward_ / self.opt_reward
else:
self.reward = self.reward_ - self.opt_reward
if self.clip_max is not None or self.clip_min is not None:
self.reward = np.clip(self.reward, self.clip_min,
self.clip_max)
if self.reward_ < int(goal):
self.reward = -1
else:
self.reward = self.reward_ / int(goal)
return self.graph, self.reward, self.done, None, self.reward_
if action in self.possible_actions:
self.enlarge_graph(action)
self.reward = self.no_reward
self.T += 1
# self.reward_ = parallel_influence(self.graph, full_graph=self.fullGraph,
# influence=self.influence_algo, times=self.times_mean)
if self.T > self.max_T:
self.done = True
# if self.reward_ < int(goal):
# self.reward = -1
# else:
# self.reward = self.reward_/int(goal)
self.reward_ = -1
return self.graph, self.reward, self.done, None, self.reward_
if action in self.active:
self.reward = self.bad_reward0
self.reward = -100
self.T += 1
if self.T > self.max_T:
self.done = True
self.reward_ = parallel_influence(self.graph,
full_graph=self.fullGraph,
influence=self.influence_algo,
times=self.times_mean)
if self.normalize:
self.reward = self.reward_ / self.opt_reward
else:
self.reward = self.reward_ - self.opt_reward
if self.clip_max is not None or self.clip_min is not None:
self.reward = np.clip(self.reward, self.clip_min,
self.clip_max)
return self.graph, self.reward, self.done, None, self.reward_
else:
raise Exception("Wrong Action:" + str(action) +
"\nPossible Actions:" + str(self.possible_actions))
def goal_setting(self, budget):
for i in range(budget):
query_node = list(self.possible_actions)
for j in query_node:
self.enlarge_graph(j)
self.goal = parallel_influence(self.graph, full_graph=self.fullGraph,
influence=self.influence_algo, times=self.times_mean)
return self.goal
def goal_setting3(self, budget, num_query, times):
self.goal = 0
for i in range(times):
for i in range(budget):
query_node = list(self.possible_actions)
query_node = random.sample(query_node, num_query)
for j in query_node:
self.enlarge_graph(j)
tmp_goal = parallel_influence(self.graph, full_graph=self.fullGraph,
influence=self.influence_algo, times=self.times_mean)
if tmp_goal > self.goal:
self.goal = tmp_goal
return self.goal
def num_infl(self):
self.num_infl = parallel_influence(self.graph, full_graph=self.fullGraph,
influence=self.influence_algo, times=self.times_mean)
return self.num_infl