Explaining and Harnessing Adversarial Examples¶
Goodfellow, I. J., Shlens, J., & Szegedy, C. (2015). Explaining and Harnessing Adversarial Examples. International Conference on Learning Representations (ICLR).
Cited by¶
3 citations across 3 artifacts.
Each citation links to the sentence it supports in the citing article.
Primes¶
- Negative Case Analysis
- Policy evaluation: a programme theory that explains observed successes is hardened by deliberately studying the sites or cohorts where it failed. Product research and UX: systematic study of churned users, failed sales, rejected applications, and abandoned carts is negative case analysis on the implicit product theory. Machine learning: adversarial testing, red-teaming, slice analysis, and out-of-distribution evaluation seek inputs that should break the model under its current account of its domain.
This sourceEstablishes adversarial-example testing — deliberately seeking small worst-case inputs that break a model's implicit account of its domain; the ML instance of disconfirmation search.
- Policy evaluation: a programme theory that explains observed successes is hardened by deliberately studying the sites or cohorts where it failed. Product research and UX: systematic study of churned users, failed sales, rejected applications, and abandoned carts is negative case analysis on the implicit product theory. Machine learning: adversarial testing, red-teaming, slice analysis, and out-of-distribution evaluation seek inputs that should break the model under its current account of its domain.
- Vaccine Escape
- In adversarial machine learning it is the input perturbation selected to land just on the wrong side of a learned decision boundary.
This sourceAdversarial perturbations selected to land just past a learned decision boundary.
- In adversarial machine learning it is the input perturbation selected to land just on the wrong side of a learned decision boundary.
Domain-specific¶
- Input Manipulation Attack
- A deep image classifier assigns an unremarkable photo of a panda the label "panda" with 57.7% confidence
This sourceGoodfellow, Shlens and Szegedy's worked Fast Gradient Sign Method example, in which GoogLeNet labels the panda image 'panda' at 57.7% confidence before the perturbation.
SupportedVerified against a saved copy of the source
“A demonstration of fast adversarial example generation applied to GoogLeNet”
- A deep image classifier assigns an unremarkable photo of a panda the label "panda" with 57.7% confidence
Verification¶
Does it exist? Not checked yet. This entry carries no identifier to resolve. It was extracted from the citation as written in the article, normalized, and deduplicated against the rest of the registry.
Does it back the claim? Read against the text for 1 of 3 citations: 1 supported. Each verdict is shown under its citation below, with what in the work backs the sentence.
Support is checked per citation rather than per work — the same source can be cited soundly in one article and wrongly in another. Per-citation recording began recently, so a citation with no recorded check is a gap in the record rather than evidence it went unchecked.
See how references were verified.
Registry ID ref:9ca037d343dd · see in the full table