Memory Consolidation through Reinstatement in a Connectionist Model of Hippocampus and Neocortex.
Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
Current memory models assume that consolidation of long-term memory in humans is facilitated by the repeated reinstatement of previous activations in the cortex. These reactivations are known to be driven by the hippocampus as part of the medial temporal lobe memory system. It has been shown, that by implementing a Hebbian depression of synaptic connections, a special kind of biologically inspired artificial neural network called Bayesian Confidence Propagation Neural Network can autonomously reinstate previously learned attractors.
Three populations of these networks, modeling short-term memory in the prefrontal cortex, mid-term memory in the medial temporal lobe, and long-term memory in the cortex, are interlinked to show that this model can produce the necessary dynamics for successful memory consolidation.
Furthermore, the resulting learning system is shown to exhibit classical memory effects shown in experimental studies, such as retrograde and anterograde amnesia after hippocampal lesioning as well as some of the effects of sleep deprivation and dopaminergic plasticity modulation on memory consolidation.
Place, publisher, year, edition, pages
Trita-CSC-E, ISSN 1653-5715 ; 2012:071
IdentifiersURN: urn:nbn:se:kth:diva-130937OAI: oai:DiVA.org:kth-130937DiVA: diva2:654383
Master of Science - Systems, Control and Robotics
Lundqvist, MikaelBenjaminsson, Simon