Densely connected normalizing flows (CROSBI ID 715880)
Prilog sa skupa u zborniku | izvorni znanstveni rad | međunarodna recenzija
Podaci o odgovornosti
Grcić, Matej ; Grubišić, Ivan ; Šegvić, Siniša
engleski
Densely connected normalizing flows
Normalizing flows are bijective mappings between inputs and latent representations with a fully factorized distribution. They are very attractive due to exact likelihood evaluation and efficient sampling. However, their effective capacity is often insufficient since the bijectivity constraint limits the model width. We address this issue by incrementally padding intermediate representations with noise. We precondition the noise in accordance with previous invertible units, which we describe as cross-unit coupling. Our invertible glow-like modules increase the model expressivity by fusing a densely connected block with Nyström self-attention. We refer to our architecture as DenseFlow since both cross-unit and intra-module couplings rely on dense connectivity. Experiments show significant improvements due to the proposed contributions and reveal state-of-the-art density estimation under moderate computing budgets.
normalizing flows, dense connectivity
nije evidentirano
nije evidentirano
nije evidentirano
nije evidentirano
nije evidentirano
nije evidentirano
Podaci o prilogu
1-15.
2021.
objavljeno
Podaci o matičnoj publikaciji
Ranzato, Marc'Aurelio ; Beygelzimer, Alina ; Nguyen, K. ; Liang, Percy S. ; Wortman Vaughan, Jennifer ; Dauphin, Yann
online: NeurIPS
Podaci o skupu
35th Conference on Neural Information Processing Systems (NeurIPS 2021)
poster
06.12.2021-14.12.2021
Toronto, Kanada; online