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 Plus annual

Presents the access of premium content as premium feature

Premium Content

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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Privacy-preserving data release algorithms have gained increasing attentionfor their ability to protect user privacy while enabling downstream machinelearning tasks. However, the utility of current popular algorithms is notalways satisfactory. Mixup of raw data provides a new way of data augmentation,which can help improve utility. However, its performance drasticallydeteriorates when differential privacy (DP) noise is added. To address thisissue, this paper draws inspiration from the recently observed Neural Collapse(NC) phenomenon, which states that the last layer features of a neural networkconcentrate on the vertices of a simplex as Equiangular Tight Frame (ETF). Wepropose a scheme to mixup the Neural Collapse features to exploit the ETFsimplex structure and release noisy mixed features to enhance the utility ofthe released data. By using Gaussian Differential Privacy (GDP), we obtain anasymptotic rate for the optimal mixup degree. To further enhance the utilityand address the label collapse issue when the mixup degree is large, we proposea Hierarchical sampling method to stratify the mixup samples on a small numberof classes. This method remarkably improves utility when the number of classesis large. Extensive experiments demonstrate the effectiveness of our proposedmethod in protecting against attacks and improving utility. In particular, ourapproach shows significantly improved utility compared to directly trainingclassification networks with DPSGD on CIFAR100 and MiniImagenet datasets,highlighting the benefits of using privacy-preserving data release. We releasereproducible code in https://github.com/Lidonghao1996/NeuroMixGDP.

NeuroMixGDP: A Neural Collapse-Inspired Random Mixup for Private Data Release