I have read a lot about RL algorithms, that update the action-value function at each step with the currently gained reward. The requirement here is, that the reward is obtained after each step.

I have a case, where I have three steps, that have to be passed in a specific order. At each step the agent has to make a choice between a range of actions. The actions are specific for each step.

To give an example for my problem: I want the algorithm to render a sentence of three words. For the first word the agent may choose a word out of ['I', 'Trees'], the second word might be ['am', 'are'] and the last word could be choosen from ['nice', 'high']. After the agent has made its choices, the reward is obtained once for the whole sentence.

Does anyone know which algorithms to use in this kind of problem?

To give a bit more detail on what I already tried: I thought that using value-iteration would be an reasonable approach to test. My problem here is that I don't know how to assign the discounted reward for the choosen opeartions.

For example after the last choice I get a reward of 0.9. But how do I update the value for the first action (choosing out of I and Trees in my example)?

I have read a lot about RL algorithms, that update the action-value function at each step with the currently gained reward. The requirement here is, that the reward is obtained after each step.

I have a case, where I have three steps, that have to be passed in a specific order. At each step the agent has to make a choice between a range of actions. The actions are specific for each step.

To give an example for my problem: I want the algorithm to render a sentence of three words. For the first word the agent may choose a word out of ['I', 'Trees'], the second word might be ['am', 'are'] and the last word could be chosen from ['nice', 'high']. After the agent has made its choices, the reward is obtained once for the whole sentence.

Does anyone know which algorithms to use in this kind of problem?

To give a bit more detail on what I already tried:

I thought that using value iteration would be an reasonable approach to test. My problem here is that I don't know how to assign the discounted reward for the chosen operations.

For example after the last choice I get a reward of 0.9. But how do I update the value for the first action (choosing out of I and Trees in my example)?

I have read a lot about RL algorithms, that update the action-value function at each step with the currently gained reward. The requirement here is, that the reward is obtained after each step.

I have a case, where I have three steps, that have to be passed in a specific order. At each step the agent has to make a choice between a range of actions. The actions are specific for each step.

To give an example for my problem: I want the algorithm to render a sentence of three words. For the first word the agent may choose a word out of ['I', 'Trees'], the second word might be ['am', 'are'] and the last word could be choosen from ['nice', 'high']. After the agent has made its choices, the reward is obtained once for the whole sentence.

Does anyone know which algorithms to use in this kind of problem?

I have read a lot about RL algorithms, that update the action-value function at each step with the currently gained reward. The requirement here is, that the reward is obtained after each step.

I have a case, where I have three steps, that have to be passed in a specific order. At each step the agent has to make a choice between a range of actions. The actions are specific for each step.

To give an example for my problem: I want the algorithm to render a sentence of three words. For the first word the agent may choose a word out of ['I', 'Trees'], the second word might be ['am', 'are'] and the last word could be choosen from ['nice', 'high']. After the agent has made its choices, the reward is obtained once for the whole sentence.

Does anyone know which algorithms to use in this kind of problem?

To give a bit more detail on what I already tried: I thought that using value-iteration would be an reasonable approach to test. My problem here is that I don't know how to assign the discounted reward for the choosen opeartions.

For example after the last choice I get a reward of 0.9. But how do I update the value for the first action (choosing out of I and Trees in my example)?