摘要
机器人与机器学习的新闻记者兼新闻编辑每日新闻-关于智能系统的新研究结果已经发表。根据NewsRx编辑在昆士兰科技大学的新闻报道，研究表明，“对抗性攻击对深度学习模型，特别是医学图像构成了重大威胁，因为它们可以通过对输入数据引入微妙的解释，误导模型做出不准确的预测，而这些解释往往是人类无法察觉的。尽管适应性训练是减轻这些医学图像攻击的常用技术，它缺乏处理新攻击方法和有效改进特征表示的灵活性。这项研究的资助者包括昆士兰技术大学；澳大利亚N研究委员会。
Abstract
By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News – New study results on intelligent syste ms have been published. According to news reporting out of Queensland University of Technology by NewsRx editors, research stated, “Adversarial attacks pose a s ignificant threat to deep learning models, specifically medical images, as they can mislead models into making inaccurate predictions by introducing subtle dist ortions to the input data that are often imperceptible to humans. Although adver sarial training is a common technique used to mitigate these attacks on medical images, it lacks the flexibility to address new attack methods and effectively i mprove feature representation.” Funders for this research include Queensland University of Technology; Australia n Research Council.
NETL
NSTL