Michael Fauth
Wed 11 Jan 2017, 14:00 - 15:00
IF 4.31/4.33

If you have a question about this talk, please contact: Gareth Beedham (gbeedham)

Long-term information storage by the collective dynamics of multi-synaptic connections


Excitatory synapses in cortex typically reside on dendritic spines.

Although cortical synapses play an important role in long-term memory, these spines undergo a remarkably high turnover. This poses the question how information can be stored on a variable substrate as synapses. As a possible solution, we propose that information is stored and retained by the collective dynamics of multiple synapses. Such a collective dynamics can already be found on the connection between two neurons, which can consist of multiple synapses. The experimentally obtained distribution of the number of synapses on these connections, which are bimodal with peaks at zero and multiple synapses, can only emerge from a collective dynamics of the involved synapses.

  We show that such a collective dynamics can emerge from the interaction of synaptic and structural plasticity and that it can be influenced by external stimulation such that the neurons can be driven to be either unconnected or connected with multiple synapses . When investigate the information storage and retention properties of these collective dynamics we find that it enables information retention on time scales orders of magnitudes longer than the typical lifetime of a synapse. Thus, the conflict of spine turnover and long- term memory can be resolved by storing information in the collective dynamics of multiple synapses.


Yet, at different external stimulation levels, where the collective dynamics yield distributions with only a single peak either at zero or at multiple synapses, information about the initial conditions decays quickly. This, however, implies that these stimulations can be used to learn new information orders of magnitude faster than it is forgotten under the conditions described above. This is  also a necessary prerequisite to solve the plasticity- stability dilemma in learning and memory on the time scale of structural changes.