Ikibazo gihuza ibibazo · Hagati Urwego

Kwibagirwa Catastrophique

Wige impamvu ukurikiranye kwiga imirimo mishya bishobora gutesha agaciro ubuhanga bwabanje nuburyo buryo bwo gukomeza-kwiga bukemura ibibazo.

Inzira zijyanye no kuyoboraKwibagirwa
Ikibazo 1 Bya 8

Nibihe bintu bisobanura kwibagirwa ibiza mumiyoboro mitsi?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. Nibihe bintu bisobanura kwibagirwa ibiza mumiyoboro mitsi?
    • Iyo wiga umurimo mushya utera icyitegererezo gutakaza ubushobozi bwize mbere
    • Iyo icyitegererezo kibuze kwibuka mugihe cyo gufata umwanzuro
    • Iyo amakuru yo guhugura asibwe kubwimpanuka
    • Iyo icyitegererezo cyanze gusubiza ikibazo
  2. Ni ukubera iki kwibagirwa ibyago bibaho mumiyoboro mitsi?
    • Kuberako buri gikorwa kibona umuyoboro wacyo mu buryo bwikora
    • Kuberako ibyiciro bishya bigezweho byanditse hejuru yuburemere busanganywe ubuhanga bukera
    • Kuberako moderi idashobora kubika ibintu birenze kimwe
    • Kuberako ibyitonderwa bisiba amakuru kubushake
  3. Niki Elastic Weight Consolidation (EWC) ikora kugirango igabanye kwibagirwa?
    • Gusiba amakuru mashya yamakuru
    • Kabiri igipimo cyo kwiga
    • Ihana impinduka kuburemere bwingenzi kubikorwa byize mbere
    • Kuraho ibitekerezo kurugero
  4. Nigute uburambe bwo gusubiramo (repetition) kurwana kwibagirwa?
    • Muguhagarika icyitegererezo cyose burundu
    • Mumyitozo gusa kumakuru mashya
    • Mugabanye umubare wibice
    • Muguhuza ingero zabitswe mugihe cyamahugurwa mashya
  5. Niki LoRA ikora kugirango ibanze-yerekana uburemere mugihe wongeyeho adaptate yihariye?
    • It freezes base weights and trains a low-rank adapter, which may reduce direct interference but does not guarantee retention
    • It rewrites every base weight for each new task
    • It stores no task-specific parameters
    • It guarantees zero forgetting on every task sequence
  6. Nigute uburyo bwo kugarura-bwongerewe uburyo bwo kwibagirwa ibiza?
    • Mugusiba ubumenyi bwa kera kugirango ubone umwanya
    • Kubika ubumenyi bushya mububiko bwo hanze aho gupima
    • Mumyitozo idahwema kurusaku rudasanzwe
    • Muguhagarika ububiko bwa moderi rwose
  7. How does Gradient Episodic Memory (GEM) constrain learning on a new task?
    • It removes old task examples from memory
    • It changes the model to a nearest-class-mean classifier
    • It adjusts the gradient so the first-order approximation does not increase loss on stored prior-task examples
    • It adds a KL penalty to every parameter regardless of task
  8. How do RAG-Token and RAG-Sequence differ in the original RAG paper?
    • RAG-Token removes retrieval while RAG-Sequence removes generation
    • RAG-Sequence uses images while RAG-Token uses audio
    • RAG-Token trains without model parameters
    • RAG-Token marginalizes retrieved documents per output token, while RAG-Sequence uses one document for a whole sequence