Model Compression

Low-Rank Factorization

Decomposition of weight matrices into products of lower-rank matrices to compress dense network layers.

← Indietro