12-06-2018 Christian Tetzlaff


The formation and organization of memory representations in adaptive, self-organized neural networks


Humans and animals perceive various environmental stimuli, which are stored as memory representations in the brain. Furthermore, dependent on the details of the stimuli, different memory representations are connected with each other enabling complex behaviors. On the neuronal level, a memory representation is associated with a strongly interconnected group of neurons (so-called Hebbian cell assembly) formed by the mechanisms of synaptic plasticity. However, it is still unknown how different variations of a stimulus are dynamically combined (classification) and associated to a specific memory representation (allocation). I will present our analysis indicating that the interaction of synaptic and homeostatic plasticity across various time scales is necessary to form a synaptic structure which enables the noise-robust classification of various stimuli. In addition, synaptic and homeostatic plasticity, acting together in parallel at various synapses, yields the self-organized association of classified stimuli with a newly formed memory representation. Thereby, our theoretical analysis shows that a set of bifurcations mainly determines this allocation process in a way as recent experimental results indicate. Furthermore, given the formation of several memory representations, we show that they can be associated, discriminated, or can form a sequence, only if a multitude of synaptic and homeostatic plasticity processes determines the synaptic dynamics. In summary, our results provide further steps in understanding how the cognitive dynamics of memories emerge from the underlying, self-organized processes of synaptic and homeostatic plasticity.