Task 4: Safety Shield Training In this task, you will train a multi-class neural network classifier to predict risk levels for state action pairs. The dataset created in [Task 3] contains all necessary features and labels in a ready-to-use format.
COMP3411/COMP9814 Artificial IntelligenceAssignment 2: Safe Interactive Reinforcement LearningTerm 3, 2025Due: Friday, 14 November 2025, 5:00 PM AESTWorth: 21 marks + 4 marks tutorial participation (25% of final grade)1 IntroductionThis assignment explores the critical challenge of safe reinforcement learning with human safety interventions. As reinforcement learning agents are increasingly deployed in safety-critical applications, from autonomous vehicles … Read more
