Associative Memory-based Replay for CL

Associative memory is the brain’s ability to link related concepts or events, where the hippocampus plays a key role in storing and retrieving these associations. It is crucial for learning, as it helps create meaningful connections between new and existing knowledge. John Hopfield (now a Nobel laureate!) proposed the Hopfield network, an energy-based model of associative memory. Leveraging Hopfield network-based, associative memory replay for continual learning holds significant neuroscientific implications but remains underexplored—this was the focus of my internship project. (Note: This is a work in-progress which we plan to submit in ICML 2025!)
See my slides for an overview of Hopfield networks!