How can I load weights if I would like to continue learning? I try to use "agent.load_model('dueling_dqn_trained.h5')" where contains data from previous training. I get error: "TypeError: load_model() takes 1 positional argument but 2 were given" Very nice tutorial, Thank You
Hi Dr. Phil, I'm a Ph.D. student and I’m really grateful beyond measure for your such great work that did tremendous support to my Ph.D.! I'm really looking for a multi-agent DQN, if that is possible for you to be offered in youtube, please
Hi, Phil Thank you for you videos, great job Could you please show how to create custom environment, for example chess or any other board game with 2 or more agents. Could you please show how implement action space in board games like chess, and how to develop and train agents against each other Thank you, Phil and good luck
@@MachineLearningwithPhil Thank you for the reply, I've just finished implementing your code with a few modifications and it looks like the Dueling agent learns faster than the standard DDQN... Very nice!
@Machine Learning with Phil Can you also make a duelling deep q network using convnets in keras. I am getting a few errors. I tried making on my own but I have failed in it. I would like to see your code if you release it.
@@MachineLearningwithPhil Totally understand. Well I hope to see a video on PG w/ TF2 in the future. Thanks again for your videos and effort to teach others RL!
also doing your udemy course, thanks mate. 2 stupid questions. Is the replay buffer same thing as episodic memory? Or is it the same as a LSTM? I'm trying to understand agent 57. I saw someone saying that episodic memory is basically two nested GRU's - is that what the replay buffer is?
You're mixing concepts. The replay buffer just keeps track of the agent's experience so it can learn from it later. Some people use RNN/LSTM in their network architecture, but I haven't done so.
@@MachineLearningwithPhil I though you were using Windows. Yes I am! But right now I am working from home using Windows and I am missing the regular vim.
Sorry for commenting again Phil. Tried contacting you via email (not sure if you saw it) - if you're open for a podcast about your work and exposing your expertise to the world, would be happy to have you in one of the next episodes :) (not trying to spam here btw :D) Thanks in advance mate!