Machine learning

Destilacija znanja

Destilacija znanja je tehnika kompresije modela, koju su 2015. uveli Geoffrey Hinton i suradnici, a koja trenira mali studentski model koristeći izlaze s mekim oznakama velikog učiteljskog modela. Destilirani modeli poput DistilBERTa i TinyBERTa postižu otprilike 97% performansi većeg modela, dok rade znatno brže.

Otvorite u MethodMindUskoroVideoUskoroDownload slides

Pročitajte cijelu metodu

Samo za članove

Prijavite se besplatnim računom kako biste pročitali ovaj odjeljak.

Prijavite se

Method map

The neighbourhood of related methods — select a node to explore.

+3 more

Izvori

  1. Hinton, G., Vinyals, O. & Dean, J. (2015). Distilling the Knowledge in a Neural Network. NeurIPS Deep Learning Workshop. link
  2. Sanh, V., Debut, L., Chaumond, J. & Wolf, T. (2019). DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter. arXiv:1910.01108. link

Kako citirati ovu stranicu

ScholarGate. (2026, June 1). Knowledge Distillation (Teacher–Student Model Compression). ScholarGate. https://scholargate.app/hr/deep-learning/knowledge-distillation

Which method?

Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.

Compare side by side

Citirana u

ScholarGateKnowledge Distillation (Knowledge Distillation (Teacher–Student Model Compression)). Preuzeto 2026-06-15 s https://scholargate.app/hr/deep-learning/knowledge-distillation · Skup podataka: https://doi.org/10.5281/zenodo.20539026