Idanwo ti o sopọ mọ itọsọna · Alabọde Ipele

Idanwo Ẹkọ Imudara Aṣoju Olona-Aṣoju

Study how several reinforcement-learning agents learn under shared rewards, changing opponents, partial views and coordination constraints.

Jẹmọ itọsọna awọn ọnaẸkọ Imudara Aṣoju pupọ
Ibeere 1 ti 8

Kini o jẹ ki ayika naa jẹ 'ti kii ṣe iduro' lati irisi aṣoju kan ni MARL?

Awotẹlẹ gbogbo ibeere8 ibeere

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

  1. Kini o jẹ ki ayika naa jẹ 'ti kii ṣe iduro' lati irisi aṣoju kan ni MARL?
    • Awọn aṣoju miiran n kọ ẹkọ nigbakanna ati yiyipada awọn eto imulo tiwọn
    • Awọn ere iṣẹ ti wa ni laileto tun kọọkan isele
    • Aaye ipinle n dagba laisi ihamọ lori akoko
    • Awọn sensọ pada awọn akiyesi ariwo
  2. Kini 'ikẹkọ ti aarin pẹlu ipaniyan isọdọtun' (CTDE) tumọ si?
    • Aṣoju ti wa ni oṣiṣẹ ati ṣiṣe awọn lori kan nikan aringbungbun server
    • Alariwisi pinpin nlo alaye agbaye lakoko ikẹkọ, ṣugbọn awọn aṣoju ṣiṣẹ lori awọn akiyesi agbegbe ni imuṣiṣẹ
    • Aṣoju ọga kan n funni ni aṣẹ si gbogbo awọn miiran ni akoko asiko
    • Awọn data ikẹkọ ti wa ni aarin ṣugbọn awọn ere jẹ ipinpinpin
  3. Iru ilana mathematiki wo ni o ṣe akopọ MDP aṣoju-ọkan si awọn aṣoju pupọ?
    • Eto laini
    • A farasin Markov awoṣe
    • Ere Markov kan (ere sitochastic)
    • Automon ipinlẹ ti o ni opin
  4. Ninu eto MARL ti o ni ifọwọsowọpọ, bawo ni ẹsan ṣe jẹ eto deede?
    • Olukuluku oluranlowo maximizes ere ti o jẹ odi ti elomiran
    • Nikan aṣoju ti o yara ju gba eyikeyi ere
    • Awọn ere ti wa ni sọtọ laileto igbese kọọkan
    • Awọn aṣoju pin ere ẹgbẹ ti o wọpọ ti wọn pọ si ni apapọ
  5. Which training approach did OpenAI Five use to improve at Dota 2?
    • Self-play against copies of its own current and past policies
    • A fixed lookup table of human moves
    • A supervised-only classifier trained on image labels
    • A single policy trained only against a static scripted opponent
  6. Kini idi ti 'iṣẹ iyansilẹ' jẹ lile paapaa ni ifowosowopo MARL?
    • Awọn ere de lesekese ati pe o rọrun lati wa kakiri
    • O nira lati pinnu iru awọn iṣe aṣoju wo ni o ṣe alabapin si abajade ẹgbẹ ti o pin
    • Oluranlowo nikan lo wa ti n gba ere
    • Awọn aṣoju ko le ṣe akiyesi eyikeyi ere rara
  7. What structural constraint does QMIX place on its mixing network?
    • It must be a linear sum with fixed weights
    • It ignores all individual agent values
    • The joint action-value is monotonic in each agent’s individual value
    • It requires communication at every action step
  8. How does Value Decomposition Networks (VDN) form the team action-value?
    • By multiplying all rewards by the number of agents
    • By selecting the largest local reward only
    • By training a separate team for each agent
    • By summing the individual agents’ action-values