Skip to main content
Artificial Intelligence

Return to Question

added 107 characters in body
Source Link
Antonio Sannia
Antonio Sannia
  • 43
  • 4

In the classical examples of deep q-learning, I often see neural networks in which the input represents the state of the agent, while the output is a tuple with all the values of $Q(s, a)$ predicted for all the possible $N$ actions.

Would it be cheaper to have $N$ neural networks with a single real-valued output, one for each of the $N$ actions?

With cheaper I mean cheaper in terms of the time complexity of a single training step of the network.

In the classical examples of deep q-learning, I often see neural networks in which the input represents the state of the agent, while the output is a tuple with all the values of $Q(s, a)$ predicted for all the possible $N$ actions.

Would it be cheaper to have $N$ neural networks with a single real-valued output, one for each of the $N$ actions?

In the classical examples of deep q-learning, I often see neural networks in which the input represents the state of the agent, while the output is a tuple with all the values of $Q(s, a)$ predicted for all the possible $N$ actions.

Would it be cheaper to have $N$ neural networks with a single real-valued output, one for each of the $N$ actions?

With cheaper I mean cheaper in terms of the time complexity of a single training step of the network.

added 7 characters in body; edited tags; edited title
Source Link
nbro
nbro
  • 41.4k
  • 12
  • 114
  • 205

Is a deep q learning network In DQN, would it be cheaper to have $N$ neural networks with a single neuronreal-valued output used in some cases, one for each of the $N$ actions?

In the classical examples of deep q learing networks-learning, I often see neural networks in which the input represents the actual state of the agent when, while the output is a tuple with all the values of Q(s,a)$Q(s, a)$ predicted for all the possiblespossible $N$ actions. With N actions using N differents

Would it be cheaper to have $N$ neural networks with a single outputs is in some cases cheaper thanreal-valued output, one for each of the standard case$N$ actions?

Is a deep q learning network with a single neuron output used in some cases?

In the classical examples of deep q learing networks I often see networks in which the input represents the actual state of the agent when the output is a tuple with all the values of Q(s,a) predicted for all the possibles actions. With N actions using N differents networks with single outputs is in some cases cheaper than the standard case?

In DQN, would it be cheaper to have $N$ neural networks with a single real-valued output, one for each of the $N$ actions?

In the classical examples of deep q-learning, I often see neural networks in which the input represents the state of the agent, while the output is a tuple with all the values of $Q(s, a)$ predicted for all the possible $N$ actions.

Would it be cheaper to have $N$ neural networks with a single real-valued output, one for each of the $N$ actions?

Source Link
Antonio Sannia
Antonio Sannia
  • 43
  • 4

Is a deep q learning network with a single neuron output used in some cases?

In the classical examples of deep q learing networks I often see networks in which the input represents the actual state of the agent when the output is a tuple with all the values of Q(s,a) predicted for all the possibles actions. With N actions using N differents networks with single outputs is in some cases cheaper than the standard case?