Machines Learning - Towards a New Synthetic Autobiographical Memory

摘要: Autobiographical memory is the organisation of episodes and contextual information from an individual’s experiences into a coherent narrative, which key to our sense self. Formation recall autobiographical memories essential for effective, adaptive behaviour in world, providing necessary planning actions functions such as event reconstruction. A synthetic system would endow intelligent robotic agents with many components cognition through active compression storage historical sensorimotor data easily addressable manner. Current approaches neither fulfil these functional requirements, nor build upon recent understanding predictive coding, deep learning, neurobiology memory. This position paper highlights desiderata modern implementation based on human episodic memory, proposes that recently developed model hippocampal could be extended generalised Initial will targeted at social interaction, where current systems have had success.