search

jeshuren / rl-cotraining

Implementation of the Reinforced Co-Training paper(https://arxiv.org/pdf/1804.06035.pdf)
8 stars 2 forks source link

hi, have you finished the code 'rl-cotraining'? #1

Open shuxjweb opened 5 years ago

shuxjweb commented 5 years ago

from rl.agents.dqn import DQNAgent from rl.policy import EpsGreedyQPolicy from rl.memory import SequentialMemory

I cannot find the code 'rl'. on the other hand, is it an small error?

labels2 = np.argmax(self.model2.predict(self.view2_unlabelled[self.members[action]]), axis=-1) self.view1_train_X = np.concatenate((self.view1_train_X, self.view1_unlabelled[self.members[action]])) self.view1_train_y = np.concatenate((self.view1_train_y, to_categorical(labels2, num_classes=2)))

modified to: self.view1_train_X = np.concatenate((self.view1_train_X, self.view2_unlabelled[self.members[action]]))