Portal de Periódicos da CAPES

Threshold devices: Fractal noise and neural talk

1994; American Physical Society; Volume: 50; Issue: 4 Linguagem: Inglês

10.1103/physreve.50.2513

1538-4519

Peter Jung,

Neural Networks and Applications

We consider the statistical properties of random pulse trains generated by noisy signals imposed on a threshold device---a simple model for the information processing of a single neuron. It is shown that Markovian noise generates self-similar bursts characterized by algebraic decaying correlations and power spectra. It is further shown that the role of noise is ambiguous. For subthreshold signals, noise can enhance the performance of the threshold device, whereas above threshold noise always degrades a signal.