10
$\begingroup$

We have a tag for but it has no wiki and I'm not really sure what it is. Can anyone shed light on this?

| |
$\endgroup$
  • 6
    $\begingroup$ Deep reinforcement learning, I assume? The approach that beats people in Go, etc. $\endgroup$ – amoeba Feb 7 '18 at 7:53
  • $\begingroup$ The tag was created here: Why does Q-Learning use epsilon-greedy during testing? That supports @amoeba's guess. Another issue is what to do about the tag. $\endgroup$ – gung - Reinstate Monica Feb 7 '18 at 12:17
  • 5
    $\begingroup$ "Deep RL" is a standard term in the deep RL community, so I'm pretty sure that's what it is. I can write an excerpt, but we can also replace all 9 instances with [deep-learning] + [reinforcement-learning]. @gung $\endgroup$ – amoeba Feb 7 '18 at 12:28
  • 6
    $\begingroup$ @amoeba Thanks for doing the legwork! But there is a touch of tautology to the statement "'Deep RL' is a standard term in the deep RL community." ;-) $\endgroup$ – Sycorax Feb 7 '18 at 16:08
  • $\begingroup$ Tautology of touch! $\endgroup$ – amoeba Feb 8 '18 at 13:41
  • 1
    $\begingroup$ (Zero hits on google? Seriously? It should be a book title or something.) $\endgroup$ – amoeba Feb 8 '18 at 13:42
  • $\begingroup$ This includes stuff like deep Q networks, actor critic models, hybrid CNN-RL methods, but maybe not things like multi-armed bandits or monte carlo methods. $\endgroup$ – MachineEpsilon Feb 9 '18 at 3:29
15
$\begingroup$

Make a synonym of . If users want to emphasize that their network is "deep," they have the option to add .

| |
$\endgroup$
  • $\begingroup$ Do you want it made a synonym of [deep-learning] as well? $\endgroup$ – gung - Reinstate Monica Feb 8 '18 at 11:59
  • 2
    $\begingroup$ @gung If it all, I'd say it should rather be a synonym of [reinforcement-learning]... $\endgroup$ – amoeba Feb 8 '18 at 14:33
  • 1
    $\begingroup$ @amoeba, in 9/11, the Q already has r-l, but only 4/11 have d-l. My impression is that people start w/ the general category (ie, r-l) & then want to augment that by noting that this is the 'deep' version. That's the reasoning behind my suggestion. That said, I'm fine w/ it if people prefer the other mapping. $\endgroup$ – gung - Reinstate Monica Feb 8 '18 at 14:50
  • 2
    $\begingroup$ I'm not convinced that we need deep-learning at all since it amounts to branding, but I'm apparently in the minority. I agree with amoeba: my view is that reinforcement-learning is the more specific designation. Additionally, OP can specify that the post is about deep-learning. $\endgroup$ – Sycorax Feb 8 '18 at 14:57
  • 1
    $\begingroup$ OK, I made it a synonym of r-l. If there are no objections after a period of time, I'll merge them. $\endgroup$ – gung - Reinstate Monica Feb 9 '18 at 14:31
  • 2
    $\begingroup$ @Sycorax we don't need [deep-learning]; it's a shame that even researchers use marketing terms (and, to make it worse, often introduce them). $\endgroup$ – Franck Dernoncourt Feb 11 '18 at 19:45
  • 1
    $\begingroup$ @Sycorax do you consider this issue closed by now? (If so, pls mark this as accepted. And maybe edit to reflect what was actually done? We did not replace [deep-rl] as per my comment, we made a synonym.) $\endgroup$ – amoeba Feb 22 '18 at 8:48

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .