On the benefits of defining vicinal distributions in latent space

Mangla, Puneet; Singh, Vedant; Havaldar, Shreyas; Balasubramanian, Vineeth N

doi:10.1016/j.patrec.2021.10.016

Cited by 2 publications

(1 citation statement)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In contrast to transformations made directly to the observed feature space, another approach applies noise to learned representations of the observations, either using stochastic network architectures for the classifier f θ (Huang et al, 2016; Srivastava et al, 2014) or by perturbing observations in a learned latent space before mapping them back to the observed feature space (Liu et al, 2018; Mangla et al, 2020; Yaguchi et al, 2019).…”

Section: Structural Regularizationmentioning

confidence: 99%

Epsilon Consistent Mixup: Structural regularization with an adaptive consistency‐interpolation trade‐off

Pisztora

Huang

et al. 2022

Stat

View full text Add to dashboard Cite

In this paper, we propose Epsilon Consistent Mixup (∈mu). ∈mu is a data‐based structural regularization technique that combines Mixup's linear interpolation with consistency regularization in the Mixup direction, by compelling a simple adaptive tradeoff between the two. This learnable combination of consistency and interpolation induces a more flexible structure on the evolution of the response across the feature space and is shown to improve semi‐supervised classification accuracy on the SVHN and CIFAR10 benchmark datasets, yielding the largest gains in the most challenging low label‐availability scenarios. Empirical studies comparing ∈mu and Mixup are presented and provide insight into the mechanisms behind ∈mu's effectiveness. In particular, ∈mu is found to produce more accurate synthetic labels and more confident predictions than Mixup.

show abstract

Section: Structural Regularizationmentioning

confidence: 99%