Maswali ya Kuendelea Kujifunza na Janga la Kusahau

Majibu na maelezo huonekana unapojibu maswali.

1. 'Kusahau kwa janga' ni nini?
   • Wakati wa kujifunza kazi mpya hudhoofisha sana utendaji wa kazi zilizojifunza hapo awali
   • Wakati mfano unaishiwa na kumbukumbu na huanguka
   • Wakati data ya mafunzo inafutwa kwa bahati mbaya
   • Wakati mfano anakataa kufanya utabiri
2. Ubadilishanaji katika moyo wa kujifunza kila mara mara nyingi huitwa:
   • Biashara ya upendeleo-tofauti
   • Utulivu-plastiki shida
   • Ubadilishanaji wa kasi-usahihi
   • Biashara ya utafutaji-unyonyaji
3. Je, Ujumuishaji wa Uzito wa Elastic (EWC) hufanya nini?
   • Hufuta niuroni zisizo muhimu
   • Huongeza mara mbili kiwango cha kujifunza kwa kazi mpya
   • Huadhibu mabadiliko kwa uzani muhimu kwa kazi zilizojifunza hapo awali
   • Huhifadhi kila mfano wa mafunzo ya zamani milele
4. Je, mbinu za uchezaji wa marudio hupambana vipi na kusahau?
   • Kwa kufungia mtandao mzima kabisa
   • Kwa kutumia kiwango kikubwa cha kujifunza
   • Kwa kuondoa kazi ya kupoteza
   • Kwa kuunganisha mifano ya zamani iliyohifadhiwa au iliyotolewa kwenye mafunzo mapya
5. How does Gradient Episodic Memory (GEM) constrain a gradient update while learning a new task?
   • It projects the gradient so its first-order approximation does not increase loss on stored prior-task examples
   • It replaces all prior-task data with new examples
   • It freezes every parameter for all future tasks
   • It increases weights that were unimportant to earlier tasks
6. Kwa nini urekebishaji mzuri wa kutojua juu ya kazi mpya husababisha kusahaulika?
   • GPU ina joto kupita kiasi na kuharibu kumbukumbu
   • Mteremko wa gradient husogeza uzani ulioshirikiwa kuelekea kazi mpya bila kizuizi cha kuhifadhi za zamani
   • Kazi mpya huwa na data zaidi kila wakati
   • Mfano huo hufuta uzani wa zamani kwa makusudi
7. How does iCaRL classify a test example after learning multiple classes?
   • By retrieving the single closest word in a text corpus
   • By selecting the class with the largest Fisher information
   • By comparing its feature vector with per-class means computed from stored exemplars
   • By replaying every training image at inference
8. Why is evaluating continual learning more than measuring final accuracy on the newest task?
   • New-task accuracy is the only relevant measure
   • Old tasks are removed from the evaluation
   • The model cannot be tested on multiple tasks
   • A method must also measure retention on prior tasks and its ability to learn new ones