Timeline for How is the reward in reinforcement learning different from the label in supervised learning problems?
Current License: CC BY-SA 4.0
6 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jul 8, 2020 at 10:28 | history | edited | Neil Slater | CC BY-SA 4.0 |
edited body
|
| Jul 8, 2020 at 10:27 | comment | added | Neil Slater | @Saptam: Yes, supervised learning works with noisy labels, and that is a common assumption. Most SL methods will do the correct thing of learning an expected value based on statistics of the labels in the dataset. | |
| Jul 8, 2020 at 9:45 | comment | added | Saptam | Excellent answer. Exactly what I was looking for. Indeed, this indirect dependency (which I was missing) is what differentiates them. On a side note, from what I've read the reward distribution can be noisy. Is it possible in supervised learning that the labels themselves are less reliable? We usually take it for granted that the labels are produced by some almighty oracle. | |
| Jul 8, 2020 at 9:31 | vote | accept | Saptam | ||
| Jul 8, 2020 at 9:30 | vote | accept | Saptam | ||
| Jul 8, 2020 at 9:31 | |||||
| Jul 8, 2020 at 8:29 | history | answered | Neil Slater | CC BY-SA 4.0 |