Guide-yakabatana mibvunzo · Zvakaoma Nhanho

HuBERT Anozvitarisira Pakutaura Mibvunzo

Tarisa kuti HuBERT inogadzira sei zvibodzwa zvakabatanidzwa zvemasikirwo enzwi, iterates mavara, uye inoongorora mamiririro akadzidzwa pamabhenji ekutaura.

Nzira dzekutungamira dzinoenderanaHubert Self Supervised Speech
Mubvunzo 1 ye 8

HuBERT inoburitsa sei zvinangwa zvainoedza kufanotaura panguva yekudzidziswa?

Tarisa mubvunzo wese8 mibvunzo

Mhinduro uye tsananguro zvinoonekwa paunenge uchitora bvunzo.

  1. HuBERT inoburitsa sei zvinangwa zvainoedza kufanotaura panguva yekudzidziswa?
    • Offline k-zvinoreva kuunganidzwa kwezvinhu zvekuteerera muzvikamu zvakavanzika
    • Zvinyorwa zvevanhu
    • Random noise labels
    • Kusiyanisa ne distractor samples
  2. Ndeipi inozivikanwa chinyorwa modhi yakafuridzira HuBERT's masked-prediction training scheme?
    • GPT-3
    • BERT
    • ResNet
    • Word2Vec
  3. HuBERT inonatsiridza sei mavara ayo anotarirwa pamusoro peinotevedzana dzidziso iterations?
    • Inohaya mamwe annotators
    • Inowedzera ruzha rukuru
    • Iyo inounganidza zvakare uchishandisa iyo modhi yakagadziridzwa yemukati inomiririra
    • Inoomesa masumbu ekutanga zvachose
  4. Ndeapi maficha anowanzo kuunganidzwa muHuBERT yekutanga iteration?
    • Mazwi ekupedzisira akanyorwa
    • Mifananidzo pixels
    • Mutauri ID embeddings
    • Yakareruka MFCC acoustic maficha
  5. Nei HuBERT ichigona kudzidza chiyamuro chinobatsira kunyangwe kuungana kwayo kusina kukwana?
    • Inongoda kufanotaura maID ID panzvimbo dzakafukidzwa, izvo zvinoramba zviri basa rinoenderana
    • Iyo inofuratira masumbu zvachose
    • Inoshandisa yakakwana oracle clustering
    • Inobata nemusoro test set
  6. Which pretraining challenge does HuBERT address by using offline cluster labels?
    • Speech has no temporal structure
    • There is no lexicon of sound units available during pretraining
    • Every speaker uses the same pitch
    • Audio contains no repeated patterns
  7. In the HuBERT paper, what relative WER reductions did the 1B-parameter model report on dev-other and test-other?
    • Exactly 50% on both sets
    • No measurable change on either set
    • Up to 19% on dev-other and 13% on test-other
    • Up to 13% on dev-other and 19% on test-other
  8. Which benchmark sets did the original HuBERT paper use for speech-recognition results?
    • ImageNet uye COCO
    • GLUE and SuperGLUE
    • SQuAD and TriviaQA
    • LibriSpeech and Libri-light