Problem of coding RL - Githubissues

hellbell / ADNet

Action-Decision Networks for Visual Tracking with Deep Reinforcement Learning (CVPR 2017)

Apache License 2.0

102 stars 40 forks source link

Problem of coding RL #15

Open zhongjunping opened 4 years ago

zhongjunping commented 4 years ago

Thanks for your great work! I'm confused about the training of RL, in the code it seems you use the "accumulate_gradients_dagnn.m" to update parameters for RL, which is the same with SL stage. However, in the paper you stated that use "stochastic gradient ascent" to maximize tracking score. Can you help to explain it? Appreciate for the help!