Katharine Jarmul keynotes on common myths around privacy and security in AI and explores what the realities are, covering design patterns that help build more secure, more private AI systems.

Abstract: Class-incremental fault diagnosis requires a model to adapt to new fault classes while retaining previous knowledge. However, limited research exists for imbalanced and long-tailed data.