Neuroscience Lecture by Misha Tsodyks (Weizmann Institute of Science, Rehovot, Israel)

Title: "Information storage and retrieval in neural network models of long-term memory."

Misha Tsodyks

Time and venue: 11.00 a.m. at the Lecture Hall (room 0.10 of the Max Planck Institute of Biophysics, Max-von-Laue-Str. 3, 60438 Frankfurt am Main, Campus Riedberg)

Abstract: The dominant theoretical framework for long-term memory is attractor neural networks (ANN) in which information is encoded by neuronal ensembles and stored by Hebbian synaptic modifications. I will present two projects addressing the issues of storage and retrieval in this framework. In the first project, I will show how memory representations exhibit slow dynamics when similar input patterns are repeatedly presented to the network. Under certain conditions, memory representations collapse into single attractor encode groups of memories. This collapse can be observed in psychophysical experiments with sequences of morphed faces. In the second project, I will  address the issue of memory recall in the absence memory-specific retrieval cues, such as in free recall experiments. I will develop a an associative model of recall where each retrieved memory item is triggering the recall of the next item. This model can be cast in the language of random graph theory and some universal laws of recall can be derived that broadly account for classical free recall experiments.

Host: Andres Laan



Public Relations

Dr. Arjan Vink
T: +49 69 850033-2900
M: +49 175 2647988