Beschreibung:Neurons in the brain learn differently from neurons in artificial neural networks (ANNs): there is arguably no gradient-based learning in the brain. However, it still learns fast and efficient. But how is this possible? A sufficient part of learning in the brain is implemented as local changes in connections between neurons. We call this synaptic plasticity. Synaptic plasticity as a learning mechanism is commonly used in spiking neural networks (SNNs) – a class of neural networks similar to the real brain.
In this seminar, we will work with synaptic plasticity in SNNs. We will learn about biologically-plausible learning mechanisms for spiking networks, such as: STDP, reward-modulated STDP, BCM, intrinsic plasticity, homeostatic plasticity, etc. We will engage in programming of spiking neural networks in Python, with the focus on these and similar learning mechanisms.
The seminar is aimed at master and higher semester bachelor students who are interested to know how the brain learns, and how to model it. Intermediate knowledge of Python is required, any knowledge about spiking networks would be ideal but not required.
