ABSTRACT: Can the neural networks be tested and checked in the similar fashion in which we test software and various applications. In this project we have presented and demonstrated a neural network dependability kit which is a toolbox to check whether the neural network used for autonomous driving system is safe or not. We have followed the step by step approach in order to judge the reliability of neural networks, the toolkit covers these scientific approaches namely a) dependability metrics to check the uncertainties in the project b) checking any undesired behavior and c) runtime monitoring to check if decision made by a neural network is corresponding to the trained data.
shoaibalisaifi / nn-dependability-kit-master-swd2021 Goto Github PK
View Code? Open in Web Editor NEWTesting the dependability of a Neural Network specifically designed for the Safety Critical Autonomous Driving Systems
License: GNU Affero General Public License v3.0