New algorithm MiniFool creates physics-based adversarial attacks for neural networks

Researchers have developed MiniFool, a novel algorithm designed to create physics-inspired adversarial attacks for neural networks. This method is particularly useful for classification tasks in particle and astroparticle physics, and has been demonstrated on datasets including MNIST and data from the CMS experiment at the Large Hadron Collider. MiniFool works by minimizing a cost function that balances a chi-squared test statistic with the deviation from a target score, allowing for the quantification of network robustness against experimental uncertainties. AI

IMPACT This research could lead to more robust neural network models in scientific applications by identifying vulnerabilities through physics-constrained adversarial attacks.

RANK_REASON The cluster contains a research paper detailing a new algorithm for adversarial attacks on neural networks, published on arXiv. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• CMS experiment
• Deep Neural Networks
• IceCube Neutrino Observatory
• Large Hadron Collider
• MiniFool
• MNIST

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →