2005 Special issue: Methods for reducing interference in the Complementary Learning Systems model: Oscillating inhibition and autonomous memory rehearsal

摘要: The stability-plasticity problem (i.e. how the brain incorporates new information into its model of world, while at same time preserving existing knowledge) has been forefront computational memory research for several decades. In this paper, we critically evaluate well Complementary Learning Systems theory hippocampo-cortical interactions addresses problem. We identify two major challenges model: Finding a learning algorithm cortex and hippocampus that enacts selective strengthening weak memories, punishment competing memories; preventing catastrophic forgetting in case non-stationary environments when items are temporarily removed from training set). then discuss potential solutions to these problems: First, describe recently developed leverages neural oscillations find parts memories (so they can be strengthened) strong competitors punished), show outperforms other algorithms (CPCA Hebbian Leabra memorizing overlapping patterns. Second, autonomous re-activation (separately hippocampus) during REM sleep, coupled with oscillating algorithm, reduce rate input patterns no longer present environment. simple demonstration process prevent interference an AB-AC paradigm. . ate