首页> 外国专利> A method for generating a test pattern for verifying the integrity of a convolution operation, selecting an optimized test pattern from the test patterns, and improving a defect tolerance capability and a swing-resistant property in an extreme situation, and apparatus

A method for generating a test pattern for verifying the integrity of a convolution operation, selecting an optimized test pattern from the test patterns, and improving a defect tolerance capability and a swing-resistant property in an extreme situation, and apparatus

机译:生成用于验证卷积运算的完整性的测试图案,从测试图案中选择优化的测试图案并在极端情况下提高缺陷容忍能力和耐摆动性的方法和装置

摘要

PROBLEM TO BE SOLVED: To verify the integrity of one or more parameters of a convolutional neural network (CNN) using at least one test pattern, which enables an autonomous vehicle to safely drive from the threat of hackers. Provide a method of generating one or more test patterns. A method uses a predetermined function to generate a test pattern so that saturation does not occur in a process in which a convolution operation is applied after the test pattern is input to at least one original CNN. Generate an evaluation score of each test pattern by referring to each test pattern and one or more parameters of the original CNN, and refer to each evaluation score to select an optimized test pattern from the test patterns. To select. [Selection diagram]
机译:要解决的问题:使用至少一种测试模式来验证卷积神经网络(CNN)的一个或多个参数的完整性,从而使自动驾驶汽车能够安全地免受黑客威胁。提供一种生成一个或多个测试模式的方法。一种方法使用预定功能来生成测试图案,使得在将测试图案输入到至少一个原始CNN之后进行卷积运算的过程中不会发生饱和。通过参考每个测试图案和原始CNN的一个或多个参数来生成每个测试图案的评估分数,并参考每个评估分数从测试图案中选择优化的测试图案。选择。 [选择图]

著录项

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号