CVPR 2020

Samareh Samadi

July 15, 2019

Domain Adaptation

Review

  • Deep Visual Unsupervised Domain Adaptation for Classification Tasks: A Survey

DFA

  • Discriminative Feature Alignment: Improving Transferability of Unsupervised Domain Adaptation by Gaussian-guided Latent Alignment Idea

Network

Step A

  • Train Genrator and Classifiers

Step B

Step C

Algorithm

Performance

JoCoR

  • Combating Noisy Labels by Agreement: A Joint Training Method with Co-Regularization

Algorithm

ADA

  • Learning to Learn Single Domain Generalization

ADA

\[x^+_{t+1} \rightarrow x^+_t +\gamma \Delta_{x^+_t} L_{ADA}(θ, ψ; x^+_t, z^+_t)\] \[L_{ADA}=L_{task}(\theta;\bf{x})-\alpha L_{const}(\theta;\bf{z})+ \beta L_{relax}(\psi;\bf{x})\]

\[L_{task}(y,\hat{y})=- \sum_i y_i log\hat{y}_i\]

\[L_{const}=\frac{1}{2}|z-z^+|_2^2 +\infty. \textbf{1} \{ y \neq y^+ \}\]

\[\min_\psi [|G(Q(\bf{x})) − x|^2 + \lambda D_e(Q(\bf{x}), P(\bf{e}))]\]

\[L_{relax}=| x^+ - V(x^+) |^2\]

Relax

عنوان اول

این متن فارسی باید راست به چپ نشان داده شود.