Maswali ya Kujifunza ya Kuimarisha Nje ya Mtandao

Majibu na maelezo huonekana unapojibu maswali.

1. Ni nini hufafanua mafunzo ya kuimarisha nje ya mtandao (bechi)?
   • Kujifunza kutoka kwa mkusanyiko wa data usiobadilika bila mwingiliano mpya wa mazingira
   • Kujifunza kupitia uchunguzi endelevu wa moja kwa moja pekee
   • Mafunzo bila ishara yoyote ya malipo
   • Kuendesha makisio bila modeli yoyote
2. Je, changamoto kuu ya kiufundi katika RL ya nje ya mtandao ni ipi?
   • Ukosefu wa data yoyote
   • Kuhama kwa usambazaji na kukadiria kupita kiasi kwa vitendo vya nje ya usambazaji
   • Ugunduzi mwingi wa mtandaoni
   • Vigezo vichache sana kwenye mtandao
3. What does D4RL provide to researchers evaluating offline reinforcement-learning algorithms?
   • A guarantee that an offline policy is safe to deploy
   • A new optimizer that removes the need for logged data
   • A benchmark suite of datasets and tasks for offline RL evaluation
   • A dataset containing only online interactions collected by each test policy
4. Je, Mafunzo ya Kihafidhina ya Q-Learning (CQL) yanapambana vipi na ukadiriaji kupita kiasi?
   • Kwa kupandisha thamani kwa vitendo visivyoonekana
   • Kwa kuondoa thawabu kabisa
   • Kwa kuchunguza zaidi katika ulimwengu wa kweli
   • Kwa kuhalalisha thamani za Q kwa vitendo vya nje ya usambazaji
5. Je, Kibadilisha Uamuzi kinawekaje upya muundo wa RL nje ya mtandao?
   • Kama uundaji wa mfuatano ambao hutabiri vitendo vilivyowekwa kwa kurudi unayotaka
   • Kama mkusanyiko wa majimbo
   • Kama utafutaji wa nasibu safi
   • Kama uainishaji wa picha unaosimamiwa
6. Kwa nini algoriti za kawaida za mtandaoni za RL haziwezi kutumika kama zilivyo nje ya mtandao?
   • Hawana mtandao wa neva
   • Hawawezi kukusanya data ya kurekebisha ili kurekebisha ukadiriaji wa thamani kupita kiasi
   • Kamwe hawatumii tuzo
   • Hawahitaji sera
7. How does BCQ reduce reliance on actions absent from an offline dataset?
   • It adds random actions to every training transition
   • It assumes every action has been observed equally often
   • It constrains candidate actions to stay near actions the dataset supports
   • It removes the action-value function and reward
8. What value-learning method is central to the cited Implicit Q-Learning paper?
   • Maximum-likelihood fitting of a language tokenizer
   • Uniform regression over actions sampled from a live environment
   • A supervised image-classification loss
   • Expectile regression over values associated with dataset actions