Event-based Learning and Computation

Networks of neurons process information by sending and receiving spikes in an asynchronous, distributed, and adaptive fashion. Biological experiments have provided us with insights into the functions of single neurons and synapses, as well as into organizational principles of neural circuits. In this project we investigate how cognitive functions can arise within spiking neural networks, and how such functions can be learned from experience. To do so we develop event-based implementations of algorithms that have been proven to be successful in machine learning and artificial intelligence, and put them to test in real-world tasks. From this approach we expect not only a deeper understanding of brain-like computation, but also develop event-based algorithms for concrete technological applications in neuromorphic systems, which have great potential for efficient real-time intelligent systems. Specifically we develop models for event-based Bayesian inference and learning through STDP, including event-based Deep Learning approaches, and event-based vision and sensory-fusion algorithms.