State and Action as Input vs State as Input and Q Values as Output

Course Q&A Machine Learning Specialization Unsupervised Learning, Recommenders, Reinforcement
1

In the course lecture, the neural network should take state and action pair as input and the target variable should be the q-function calculated using the bellman equation. If that is the case, why in Lunar Lander Exercise are we initializing the network input with only the state and not state-action pair. I am a bit confused on this.

1 Like
2

Hello @mateen

The following lecture explains that the NN does not have to take state-action pair as input.

Cheers,
Raymond

2 Likes
3

Thanks @rmwkwok

2 Likes