Oṣere-Critic Awọn ọna adanwo

Awọn idahun ati awọn alaye han bi o ṣe n mu ibeere naa.

1. Ninu ọna Oṣere-Oṣere, kini ipa ti 'alariwisi' naa?
   • Lati ṣe iṣiro iye awọn ipinlẹ (tabi awọn iṣe) ati ṣe iṣiro awọn yiyan oṣere naa
   • Lati yan iru igbese lati ṣe
   • Lati tọju ifipamọ atunṣe
   • Lati ṣawari agbegbe laileto
2. Kini 'anfani' A(s,a) = Q(s,a) -V(s) wiwọn?
   • Lapapọ ere isele
   • Elo ni iṣe ti o dara julọ ju iye apapọ ipinlẹ lọ
   • Oṣuwọn ẹkọ
   • Nọmba awọn iṣe ti o wa
3. Kini idi ti fifi alariwisi kan dinku iyatọ ti a fiwera si awọn ọna imuduro eto imulo mimọ bii REINFORCE?
   • O yọ awọn ere lati idogba
   • O mu ki eto imulo pinnu
   • Alariwisi n pese ipilẹsẹ kan, nitorinaa awọn imudojuiwọn lo ami ifihan anfani ariwo ti o kere si
   • O mu ki awọn eni ifosiwewe
4. Agbara wo ni awọn ọna oṣere-alariwisi ni awọn ọna ti o da lori iye itele bi Q-Learning mu lainidi?
   • Awọn iṣe ọtọtọ
   • Lilo awọn ere
   • Idinku ojo iwaju
   • Awọn aaye iṣe ilọsiwaju
5. Ewo ninu atẹle jẹ algorithm ti o da lori oṣere-Oṣere ti o lo pupọ fun RLHF lori awọn awoṣe ede?
   • PPO (Imudara Ilana Isunmọ)
   • K-tumo si
   • Ipadasẹyin laini
   • Q-Ẹkọ
6. Ifihan agbara wo ni oṣere maa n lo lati ṣe imudojuiwọn eto imulo rẹ?
   • Awọn aise pixel input
   • Anfani tabi aṣiṣe TD ti a pese nipasẹ alariwisi
   • Iwọn ifipamọ atunṣe
   • A ti o wa titi ID nọmba
7. An action has a positive advantage estimate. How does a policy-gradient actor typically use that signal?
   • Decrease its probability in every state
   • Set the critic’s value estimate to zero
   • Increase the probability of choosing that action in the state
   • Replace the policy with a Q-table
8. An action’s estimated advantage is negative. What direction should the actor update favor?
   • Increase it because the critic was surprised
   • Delete the state from the environment
   • Raise the reward discount factor automatically
   • Reduce the likelihood of that action relative to the policy baseline