Ikibazo gihuza ibibazo · Biragoye Urwego

Ntibisanzwe Autoencoders kubibazo byo gusobanura

Kugenzura gake autoencoder kwiyubaka no kuba gake, ibikorwa byo gukora, kwaguka kwagutse, gutandukanya ibice hamwe nubushakashatsi bwo kuyobora.

Inzira zijyanye no kuyoboraNtibisanzwe Autoencoders Kubisobanuro
Ikibazo 1 Bya 8

Niyihe ntego nyamukuru yo guhugura autoencoder idasanzwe kubikorwa bya moderi?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. Niyihe ntego nyamukuru yo guhugura autoencoder idasanzwe kubikorwa bya moderi?
    • Kubora ibikorwa bifatika mubisobanuro, ahanini-byitaruye
    • Kugirango ugabanye icyitegererezo kugirango gikore vuba
    • Guhindura inyandiko hagati yindimi
    • Kongera icyitegererezo cyukuri kubipimo
  2. How does an SAE’s sparsity penalty affect hidden-unit activations?
    • It guarantees each unit represents exactly one concept
    • It encourages only a small subset of the wider hidden layer to activate for an input
    • It makes every hidden unit activate equally
    • It removes the reconstruction objective
  3. Hafi y'ibintu bingahe Anthropic yakuyemo mugihe yapimye SAE kuri Claude 3 Sonnet muri 2024?
    • Abagera kuri 500
    • Abagera ku 2000
    • Abagera kuri miliyoni 34
    • Hafi miliyari 100
  4. Ni iki 'Irembo rya Zahabu Claude' ryerekanaga?
    • Ko SAEs ishobora gutozwa gusa kumashusho
    • Ko icyitegererezo kitagishoboye kubyara inyandiko
    • Ibyo biranga nta ngaruka bigira
    • Ibyo gufatira kumurongo umwe murwego rwo hejuru bitera imyitwarire yicyitegererezo
  5. Ni ukubera iki igicucu cyihishe muri SAE mubisanzwe CYANE kuruta ibikorwa byubaka?
    • Kuberako ibitekerezo byashyizwe mubipimo bike ugereranije nibyo bakeneye, bityo ibice byinshi bifasha kubitandukanya
    • Kugira ngo amahugurwa ahendutse
    • Kuberako ibice byagutse buri gihe birasobanutse neza
    • Kugabanya umubare wibipimo
  6. Nibihe bizwi ko bigarukira kuri autoencoders zidasanzwe?
    • Ntibashobora gutozwa hamwe no kumanuka
    • Gutandukanya ibiranga, aho igitekerezo kimwe kimeneka mubintu byinshi bitandukanye
    • Bakora gusa kumurongo wemeza
    • Bakuraho ibikenewe kuranga abantu
  7. Which two objectives are commonly combined when training a sparse autoencoder?
    • A policy gradient and value loss
    • A next-token loss and a contrastive loss
    • Reconstruction error and a penalty on hidden activations
    • A calibration error and classification accuracy
  8. What does a TopK sparse autoencoder do to encourage sparsity?
    • Keeps the K largest input pixels and discards the decoder
    • Sets all encoder weights to zero
    • Trains K different tokenizers
    • Keeps only the K largest hidden activations for an input