What proportion of tf_agents choices are random and how many are based upon past experiences?

Viewed 16

Correct me if i'm assuming something wrong.

But in other ml systems, things start off completely random. (In this case they do too , im using a random policy) But gradually , larger proportions of the actions chosen are decided upon based upon previous experience instead of randomness.

Is there something like tf_agents.proportion_choices_random , or tf_agents.policy.proportion_choices_random that will return 0 initially when all choices are random , then like 0.4 a many episodes later?

0 Answers
Related