{"componentChunkName":"component---src-templates-blog-post-js","path":"/blog/robustness-of-multi-exit-architectures-against-adversarial-attacks/","result":{"data":{"site":{"siteMetadata":{"title":"No Frills News"}},"contentfulNfnPost":{"postTitle":"Robustness Of Multi-Exit Architectures Against Adversarial Attacks","slug":"robustness-of-multi-exit-architectures-against-adversarial-attacks","createdLocal":"2021-05-30 14:30:44.587646","publishDate":"2021-05-30 11:30:00+00:00","feedName":"Image Recognition","sourceUrl":{"sourceUrl":"https://analyticsindiamag.com/robustness-of-multi-exit-architectures-against-adversarial-attacks/"},"postSummary":{"childMarkdownRemark":{"html":"<p>To remediate this, researchers have devised multi-exit architectures where deeper networks fix mistakes of shallow networks.\nIn multi-exit architectures, the stack of processing layers are interleaved with early output layers or ‘exits’.\nHowever, the computational savings provided by this approach may not be robust against adversarial attacks.\nDeepSlothRecently, a group of researchers from the University of Maryland and Alexandru Ioan Cuza University studied the robustness of multi-exit architectures against adversarial attacks; and discovered DeepSloth.\nThe standard defence against the adversarial samples was found to be inadequate against the slowdown.</p>"}}}},"pageContext":{"slug":"robustness-of-multi-exit-architectures-against-adversarial-attacks"}}}