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

Auditing trained deep learning (DL) models prior to deployment is vital forpreventing unintended consequences. One of the biggest challenges in auditingis the lack of human-interpretable specifications for the DL models that aredirectly useful to the auditor. We address this challenge through a sequence ofsemantically-aligned unit tests, where each unit test verifies whether apredefined specification (e.g., accuracy over 95%) is satisfied with respect tocontrolled and semantically aligned variations in the input space (e.g., inface recognition, the angle relative to the camera). We enable such unit teststhrough variations in a semantically-interpretable latent space of a generativemodel. Further, we conduct certified training for the DL model through a sharedlatent space representation with the generative model. With evaluations on fourdifferent datasets, covering images of chest X-rays, human faces, ImageNetclasses, and towers, we show how AuditAI allows us to obtain controlledvariations for certified training. Thus, our framework, AuditAI, bridges thegap between semantically-aligned formal verification and scalability. A blogpost accompanying the paper is at this linkhttps://developer.nvidia.com/blog/nvidia-research-auditing-ai-models-for-verified-deployment-under-semantic-specifications

Auditing AI models for Verified Deployment under Semantic Specifications