Idanwo ti o sopọ mọ itọsọna · Alabọde Ipele

Idanwo Ẹkọ Afarawe

Ṣiṣayẹwo cloning ihuwasi, iyipada pinpin, awọn ifihan iwé, akopọ data DAgger ati awọn idiwọn ni ita awọn ipinlẹ akiyesi.

Jẹmọ itọsọna awọn ọnaẸkọ Afarawe
Ibeere 1 ti 8

Ni ikẹkọ afarawe, bawo ni aṣoju ṣe gba ihuwasi rẹ?

Awotẹlẹ gbogbo ibeere8 ibeere

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

  1. Ni ikẹkọ afarawe, bawo ni aṣoju ṣe gba ihuwasi rẹ?
    • Kọ ẹkọ eto imulo nipa didakọ awọn ifihan iwé
    • Ti o pọju ere ti a ṣe ni ọwọ nipasẹ iṣawakiri laileto
    • Fifun nẹtiwọọki nkankikan lati ṣiṣẹ lori awọn ẹrọ kekere
    • Ṣiṣẹda awọn aworan lati awọn itọ ọrọ
  2. Awọn fireemu ẹda ẹda ihuwasi ihuwasi bi iru iṣoro wo?
    • Iṣjọpọ ti ko ni abojuto
    • Ẹ̀kọ́ àmójútó ti ìpínlẹ̀-si-iṣẹ́ ìyàwòrán
    • iran ọta
    • Iṣatunṣe ere
  3. Iṣoro wo ni o fa ki cloning ihuwasi kuna lori awọn ilana iṣe gigun?
    • Overfitting to ere iṣẹ
    • Pipalẹ gradients ninu kooduopo
    • Aṣiṣe idapọmọra lati iyipada pinpin
    • Igbagbe ajalu ti awọn kilasi atijọ
  4. Bawo ni DAgger algorithm ṣe koju ailera yẹn?
    • Nipa yiyọ amoye kuro ni lupu patapata
    • Nipa jijẹ ifosiwewe eni
    • Nipa fifi ariwo laileto kun si gbogbo awọn iṣe
    • Nipa bibeere alamọdaju lori awọn ipinlẹ olukọ n ṣabẹwo si gangan ati ṣajọpọ data yẹn
  5. Kini idi ti ẹkọ afarawe nigbagbogbo ṣe fẹ ju ẹkọ imuduro fun awọn iṣẹ ṣiṣe bii wiwakọ?
    • Awọn ifihan jẹ rọrun lati pese ju iṣẹ ere ti o tọ lọ
    • O ṣe onigbọwọ ihuwasi to dara julọ
    • Ko nilo eyikeyi data
    • O nṣiṣẹ yiyara ni itọkasi nikan
  6. What is a realistic limitation of behavioral cloning when the agent reaches states outside the demonstrations?
    • It is guaranteed to improve beyond the expert
    • Its behavior is not guaranteed there because those states may be out of distribution
    • It can only produce random actions
    • It automatically queries an expert at every step
  7. In one DAgger iteration, what does the learner do before requesting expert labels?
    • Deletes all earlier demonstrations
    • Trains only on randomly generated labels
    • Rolls out its current policy and records the states it actually visits
    • Replaces the expert policy permanently
  8. After the expert labels the states visited by the current policy, what happens to that data?
    • It is used only as an evaluation metric
    • It replaces the original observation space
    • It is discarded after one gradient step
    • It is aggregated with previous labeled data for retraining