Artificial Intelligence By Example

Finding the main target for the MDP function

W, the weight vector, is updated after each frame with a short-term memory of n frames. This means that in every n frames, it's memory is emptied of useless information.

The goal of the optimizer is to provide a target for the MDP function. On the first episode, since it does not have any information, the optimizer selects a random state, as shown in the following code excerpt:

#first episode is randomif(e==0):lr=random.randint(0,5)

This means that a random assembly station will be chosen on the MPD graph, which represents the six assembly sewing stations. Once this episode has been completed, the system enters a circular real-time cycle (see next section).

The second episode has the W vector to ...