Learning Temporally Encoded Patterns in Networks of Spiking Neurons

Abstract: Networks of spiking neurons are very powerful and versatile models with regard to their biological realism and their computational power. Especially units receiving temporally encoded inputs (as time differences between firing times) are very powerful since they can be used for simulating arbitrary feedforward sigmoidal neural nets. We investigate how neurons can learn a given weight-value only on the basis of time-differences between firing times.

Back to my homepage.