Kolloquien
URL zum ICS-Kalender dieses Seminars
Kirchhoff-Institut für Physik, Otto-Haxel-Hörsaal
freitags 17:15
Vorträge
16.11.2018 17:00
KIP, INF 227, Otto-Haxel-Hörsaal
Festkolloquium zum 80. Geburtstag von Prof. Dr. Heinz Horner
What happens in your brain when you learn to ride a bike or memorize a new word?
In this talk, I will discuss mathematical models of memory formation and learning.
Neurons communicate with each other in form of short electrical pulses. The pulses
travel along 'axonal' cables that ramify and form thousands of synapses with other
neurons. The strength of a synaptic connection is
the location where memories are
stored. Mathematical learning rules describe how and when these synaptic
connections change. Simulations and analysis of neural network models show how
such learning rules, located at the level of synaptic connections, can lead to global
changes in brain function via formation of memories. While the classic Hopfied model
from 1982, which for the first time linked the physics of spin glasses to the brain,
assumed that memories are pre-
learned, modern models of brain-
inspired neural
network try to understand how online learning upon exposure to new events is
possible.