Sebastian Theiler
Sebastian Theiler
The whole data flow is indeed fed to both the intial value and advantage streams. After that, there are seperate dense layers for the final calculations. The Lambda layer is...
I believe there is a slice/split operation in the paper: > Our network architecture has the same low-level convolutional structure of DQN... As shown in Figure 1, the dueling network...
I'm really sorry but I don't know if I'll be able to help, since it's been almost a year since I last worked on this and this was written for...
Thank you for spotting this! I've just made the fix in a local file, and am currently testing it, with the target-network being updated every 10,000 steps. I'll get back...
I apologize that this has taken me so long, the past few months have been hectic. I've created a new branch [dqn-fix](https://github.com/EvolvedSquid/tutorials/tree/dqn-fix) with the changes. I haven't had time to...