Upgrade to Pro — share decks privately, control downloads, hide ads and more …

EmbraceNet: A robust deep learning architecture...

daiki
July 15, 2023

EmbraceNet: A robust deep learning architecture for multimodal classification

J.-H. Choi, J.-S. Lee. EmbraceNet: A robust deep learning architecture for multimodal classification. Information Fusion, vol. 51, pp. 259-270, Nov. 2019

https://arxiv.org/abs/1904.09078

daiki

July 15, 2023
Tweet

More Decks by daiki

Other Decks in Research

Transcript

  1. EmbraceNet: A robust deep learning architecture for multimodal classification daiki

    J.-H. Choi, J.-S. Lee. EmbraceNet: A robust deep learning architecture for multimodal classification. Information Fusion, vol. 51, pp. 259-270, Nov. 2019 https://arxiv.org/abs/1904.09078
  2. EmbraceNet: Embracement Layer 確率𝒓𝒊 を用いて確率的に融合 ◼𝒓𝒊 はパラメータPの多項分布に従う (𝒓𝒊 の和は1) ➢𝒓𝒊

    = 𝑟 𝑖 1 , 𝑟 𝑖 2 , … , 𝑟 𝑖 𝑚 (𝑖 = 1, … , 𝑐) c:データ長 m: モダリティ数 ➢Ex) 通常は𝐏 = 1 m , 1 m , . . , 1 m (すべてのモダリティで同確立) 8 𝒓𝟏 = [1,0, 0] 𝒓𝟐 = [0,1, 0] 𝒓𝒄 = [0,1, 0] …