Ikibazo gihuza ibibazo · Biragoye Urwego

Gutoranya Kubi no Gusakuza Kugereranya Ikigereranyo

Reba uburyo NCE hamwe nicyitegererezo cyiza gikoresha ingero zurusaku, kugabanya igiciro cyuzuye-softmax, kandi gitandukanye nintego zabo zishoboka.

Inzira zijyanye no kuyoboraGutoranya nabi no gusakuza kugereranya
Ikibazo 1 Bya 8

Ni ikihe kibazo icyitegererezo cyiza na NCE gikemura cyane?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. Ni ikihe kibazo icyitegererezo cyiza na NCE gikemura cyane?
    • Igiciro kinini cyo kubara softmax yuzuye hejuru yumwanya munini cyane
    • Kurenza kuri datasets nto
    • Kuzimya gradients mumiyoboro isubiramo
    • Kubura amakuru yanditseho imyigire ikurikiranwa
  2. Nigute Urusaku rutandukanye rwo kugereranya rusubiramo imirimo yo kwiga?
    • Nkugusubira inyuma kumigambi ikomeza
    • Nka binini itondekanya hagati yamakuru yukuri nicyitegererezo
    • Nka guhuriza hamwe ingingo zidashyizweho ikimenyetso
    • Nkukwiga gushimangira hamwe nibihembo
  3. Niyihe moderi yakwirakwije icyitegererezo cyiza cyo kwiga gushyiramo ijambo?
    • BERT
    • ResNet
    • ijambo2vec's skip-gram moderi
    • GPT-2
  4. Ni gute icyitegererezo kibi gitandukanye na NCE yuzuye?
    • Yongeyeho amagambo asanzwe asanzwe kugirango yukuri
    • Irasaba amagambo yose kuri buri ntambwe
    • Ikoresha igihombo cyo gusubira inyuma aho kuba sigmoid
    • Igabanya NCE imvugo isanzwe, bigatuma yihuta ariko ntabwo igereranya ubucucike
  5. Ni ukubera iki ibibi byakunze gutangwa kuva kugabana unigram kuzamurwa kuri 3/4 imbaraga?
    • Kuringaniza ingaruka zamagambo kenshi kandi adasanzwe aho gutoranya gusa na frequency
    • Kwirengagiza amagambo adasanzwe
    • Gukora amagambo yose birashoboka
    • Gukuraho amagambo yo guhagarika imyitozo
  6. For a large vocabulary, how can sampled negative pairs reduce work compared with evaluating every word in a full softmax?
    • It computes an exact probability for every vocabulary item
    • The update scores a positive pair and a limited set of sampled negatives instead of normalizing over the whole vocabulary
    • It eliminates the need to score the positive pair
    • It sorts all words by frequency on each step
  7. In skip-gram negative sampling, what is treated as the positive training pair?
    • Two words sampled independently from the noise distribution
    • A word paired with an empty context
    • A word and a context word observed together in the corpus
    • A validation label paired with a model checkpoint
  8. Why is negative sampling not generally interpreted as estimating a normalized language-model probability?
    • It always sums over the full vocabulary
    • It requires exact maximum-likelihood normalization
    • It never uses negative examples
    • Its simplified binary objective drops NCE normalization terms and targets useful embeddings instead