Talk:End-to-end reinforcement learning

Merge with article on deep reinforcement learning?
The term "end-to-end reinforcement learning" is just another way to refer to "deep reinforcement learning" but deep RL is the more formal term. This article is actually better in giving examples of deep RL but the deep RL page is more informative/descriptive. I think these articles should be merged. Anair13 (talk) 19:59, 24 November 2020 (UTC)

Went ahead and merged Anair13 (talk) 02:30, 1 December 2020 (UTC)

Can I ask why this was restored? Especially without discussion, after I had started a discussion on it? I don't think there is any formal distinction between deep reinforcement learning and "end-to-end reinforcement learning", and I would challenge someone to find a citation saying that there is in order to keep this page and also to make the distinction clear on this page. They both just vaguely mean reinforcement learning with function approximation to handle raw inputs. Moreover, this page is unbalanced towards the work of one author, Katsunari Shibata (perhaps a violation of Neutral_point_of_view), at the exclusion of a lot more famous and foundational work. Anair13 (talk) 17:00, 27 October 2021 (UTC)

I apologize for restoring this page without discussion. Deep reinforcement learning refers to the use of a deep neural network. In most cases, a convolutional neural network is used with raw images as input. On the other hand, "end-to-end reinforcement learning" means that in reinforcement learning, the process being learned must be from one end (usually sensors) to the other (usually actuators). Therefore, the two are similar, but not exactly the same. However, as you said, this page is not balanced, and that should be solved. Therefore, for now, I am in favor of merging this page into deep reinforcement learning. Pioneerest (talk) 20:08, 28 October 2021 (UTC)

OK, thanks for the reply. If we are on the same page, shall we go ahead and merge it then? I had already merged the contents of this page into deep reinforcement learning, including the history parts and a note about end-to-end reinforcement learning. Is there anything else you want to add to the deep RL page? Anair13 (talk) 17:38, 30 October 2021 (UTC)

Yes, I would like to ask you to merge this page into deep reinforcement learning. I don't have anything more to add to the deep RL page at the moment. Thank you. Pioneerest (talk) 23:03, 30 October 2021 (UTC)

OK, will do! Anair13 (talk) 22:35, 31 October 2021 (UTC)