adversarial
Opposing someone in a hostile or competitive way instead of cooperating.
Characterized by opposition, conflict, or a confrontational attitude, as between opponents.
Used about a legal process where two sides argue against each other before a judge or jury.
In law, describing a system or proceeding in which opposing parties present arguments and evidence to a neutral decision-maker.
Made to trick, attack, or test a computer system, especially an AI model.
In computing and AI, designed to deceive, manipulate, attack, or test a system by exploiting weaknesses, as if by an opponent.
- Base: adversarial
adversarial approach
1confrontational method
A way of handling something by acting against others instead of working with them.
- The mayor rejected an adversarial approach and invited critics to join the planning committee.
- An adversarial approach to customer complaints can damage a brand quickly.
- The coach's adversarial approach motivated some players but discouraged others.
- He used an adversarial approach.
- An adversarial approach may make the problem worse.
adversarial attack
1system-fooling attack
A special attack that tries to trick an AI or computer system.
- A small change to the image caused an adversarial attack on the classifier.
- The researchers measured how often the model resisted adversarial attacks.
- Voice assistants can be vulnerable to adversarial attacks hidden in audio.
- The AI failed after an adversarial attack.
- They tested the model against adversarial attacks.
adversarial example
1input crafted to mislead
A test item made to trick an AI model into giving the wrong answer.
- The stop sign image looked normal to humans, but it was an adversarial example for the model.
- Adversarial examples reveal weaknesses in systems that seem highly accurate on standard tests.
- The team generated adversarial examples to improve the classifier.
- This picture is an adversarial example.
- The model made a mistake on the adversarial example.
adversarial legal system
1opposing-sides court system
A court system where two sides argue their cases and a judge or jury decides.
- England and Wales have an adversarial legal system.
- In an adversarial legal system, lawyers for each side control much of how evidence is presented.
- Supporters argue that an adversarial legal system protects defendants by allowing strong challenges to the prosecution's case.
- The country uses an adversarial legal system.
- Lawyers are important in an adversarial legal system.
adversarial machine learning
1machine learning under attack
The study of how AI models can be attacked or fooled and how to protect them.
- Adversarial machine learning has become more important as AI systems are used in security-critical settings.
- The course covers privacy, robustness, and adversarial machine learning.
- Her research in adversarial machine learning focuses on defending image classifiers.
- She studies adversarial machine learning.
- Adversarial machine learning helps protect AI systems.
adversarial process
1contest-based decision process
A process where two sides argue against each other so someone else can decide.
- The adversarial process gives both sides a chance to test the evidence.
- Some family courts try to reduce the harm caused by an adversarial process.
- Academic debate can be an adversarial process without becoming personal.
- The court uses an adversarial process.
- Both sides speak in the adversarial process.
adversarial relationship
1conflict-based relationship
A relationship in which people or groups act like opponents.
- The company has an adversarial relationship with the regulator.
- Years of disputes created an adversarial relationship between the union and management.
- The new director wants to replace the adversarial relationship with a more cooperative one.
- They have an adversarial relationship.
- We need a less adversarial relationship.
adversarial training
1training with challenging inputs
A way to train AI with difficult trick examples so it becomes harder to fool.
- Adversarial training improved the model's performance on noisy images.
- The lab used adversarial training to make the speech-recognition system more robust.
- Adversarial training can be expensive because it requires generating difficult examples.
- Adversarial training made the model stronger.
- The team used adversarial training.