Idanwo ti o sopọ mọ itọsọna · Lile Ipele

_AIU_PROTECTED_13_-Adanwo Ẹkọ

Review how meta-learning trains across tasks, how MAML adapts model initialization, and how prototype methods classify few-shot examples.

Jẹmọ itọsọna awọn ọnaMeta Ẹkọ
Ibeere 1 ti 8

Kini ikẹkọ-meta, tabi 'kikọ lati kọ ẹkọ,' ni akọkọ ifọkansi lati ṣaṣeyọri?

Awotẹlẹ gbogbo ibeere8 ibeere

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

  1. Kini ikẹkọ-meta, tabi 'kikọ lati kọ ẹkọ,' ni akọkọ ifọkansi lati ṣaṣeyọri?
    • Imudara yara si awọn iṣẹ ṣiṣe titun lati awọn apẹẹrẹ diẹ
    • Ti nṣe iranti ipilẹ data nla kan
    • Imukuro iwulo fun ikẹkọ eyikeyi
    • Compressing awọn awoṣe si iwọn odo
  2. Lakoko ikẹkọ-meta, bawo ni data ṣe n ṣe eto nigbagbogbo?
    • Bi awọn kan nikan-ṣiṣe pẹlu ko si yapa
    • Bii awọn iṣẹ ṣiṣe lọpọlọpọ, ọkọọkan pẹlu eto atilẹyin ati ṣeto ibeere kan
    • Bi aworan kan tun ṣe
    • Bi ariwo ti ko ni aami nikan
  3. Kini MAML kọ?
    • Atọka ti o wa titi ti ko yipada
    • A dataset ti odi
    • Ipilẹṣẹ awoṣe ti o le ṣe atunṣe daradara si awọn iṣẹ ṣiṣe titun ni awọn igbesẹ diẹ
    • A kodẹki funmorawon
  4. Bawo ni Awọn Nẹtiwọọki Afọwọṣe ṣe iyatọ awọn apẹẹrẹ titun?
    • Nipa ṣiṣẹda titun awọn aworan
    • Nipa ti nṣe iranti awọn iye piksẹli
    • Nipa lafaimo lairotẹlẹ
    • Nipa ifiwera wọn si awọn apẹrẹ kilasi ti o kọ ẹkọ ni aaye ifibọ
  5. Kini ipa ti lupu inu MAML?
    • Lati ṣe atunṣe awoṣe si iṣẹ-ṣiṣe kan pato pẹlu awọn igbesẹ gradient diẹ
    • Lati dapọ dataset
    • Lati pa awọn paramita rẹ
    • Lati ṣeto oṣuwọn ẹkọ si odo
  6. How does GPT-3 perform the few-shot demonstrations in the cited paper?
    • By updating all model weights after every example
    • By including examples in the prompt without gradient updates or fine-tuning
    • By storing one prototype per class
    • By running a MAML inner loop during inference
  7. In Reptile, how is the meta-initialization updated after adapting on a sampled task?
    • Replace the model with a nearest-neighbor table
    • Average class labels without updating parameters
    • Move the initialization toward the task-adapted parameters
    • Train a separate model for every example
  8. What computational shortcut distinguishes Reptile from the second-order MAML formulation?
    • It requires no gradient steps
    • It stores only class prototypes
    • It trains without tasks or task-specific data
    • It can use a first-order update without computing second-order derivatives