Timeline for When we use a neural network to approximate the Q values, is the Q target a single value?
Current License: CC BY-SA 4.0
3 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jun 13, 2020 at 10:33 | comment | added | Chukwudi | Ohh I think I get it, it’s a function approximation, so as long as the weights change , the next time the state is fed the Q values also change, it doesn’t matter about changing immediately | |
| Jun 12, 2020 at 22:18 | comment | added | David | I think you are misunderstanding what happens in Q-learning. The network is a function of the state. Imagine a normal neural network that is a function of x, when you update the weights the next time you feed the same x in it will (probably) give you a different value. There’s not really anything else that can be said unless you explain further what you don’t understand. | |
| Jun 12, 2020 at 16:33 | history | answered | David | CC BY-SA 4.0 |