Project Details
Description
Many of the brain’s higher cognitive functions rely on memory storage. Many studies in computational neuroscience have attempted to gain an understanding of the mechanisms of memory formation, preservation and decay. The models of memory networks designed so far capture important features, but fail to fully reproduce animals’ capability to quickly form memories that last for a long time. The best models, using an advanced synaptic mechanism with differing levels of plasticity, are still far from attaining human performance. I hypothesize that this discrepancy can be resolved if memories are stored in heterogeneous networks, which are characterized by different plasticity levels. Memories may be transferred between these networks, as indicated by biological evidence for memory transfer across brain areas (in particular during replay events). I plan to (goal 1) study the performance of memory systems that are divided into different network stages, each characterized by synapses that are modified at different rates, with the aim of obtaining a plausible model of fast-forming long-term memories. Secondly (goal 2), I plan to study the different modes of information transfer from one stage to another one though neural activity, similar to the one observed during replay events. For this purpose, I will design a more realistic multi-stage network with spiking neuron models as well as phenomenological models of synaptic plasticity.
Status | Finished |
---|---|
Effective start/end date | 2/1/12 → 8/31/13 |
Funding
- Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung
ASJC Scopus Subject Areas
- Artificial Intelligence
- Neuroscience(all)
- Biophysics