English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Deep neural networks (DNNs) have demonstrated promising results in variouscomplex tasks. However, current DNNs encounter challenges withover-parameterization, especially when there is limited training dataavailable. To enhance the generalization capability of DNNs, the Mixuptechnique has gained popularity. Nevertheless, it still produces suboptimaloutcomes. Inspired by the successful Sharpness-Aware Minimization (SAM)approach, which establishes a connection between the sharpness of the trainingloss landscape and model generalization, we propose a new learning frameworkcalled Generalized-Mixup, which combines the strengths of Mixup and SAM fortraining DNN models. The theoretical analysis provided demonstrates how thedeveloped G-Mix framework enhances generalization. Additionally, to furtheroptimize DNN performance with the G-Mix framework, we introduce two novelalgorithms: Binary G-Mix and Decomposed G-Mix. These algorithms partition thetraining data into two subsets based on the sharpness-sensitivity of eachexample to address the issue of "manifold intrusion" in Mixup. Both theoreticalexplanations and experimental results reveal that the proposed BG-Mix andDG-Mix algorithms further enhance model generalization across multiple datasetsand models, achieving state-of-the-art performance.

G-Mix: A Generalized Mixup Learning Framework Towards Flat Minima