Linear-nonlinear-Poisson cascade model
The Linear-nonlinear-Poisson (LNP) cascade model [1][2] is used to encode the stimuli of neural ensembles in the visual system. As opposed to traditional approaches, which use stimuli such as drifting or pulsating bars, the LNP cascade model uses white noise as the stimuli.
There are three stages of the LNP cascade model. The first stage computes a weighted sum of the stimulus vector, which is then passed to the nonlinear function of the next stage. The result of the nonlinear function then determines the firing rate of the Poisson spike generator.
The stimulus vector is a vector that contains the spatiotemporal receptive field. For example, if there are four pixels in the stimulus and six time steps are used, then the stimulus vector would contain twenty-four elements. The weighted vector used in the first stage can be calculated by reverse correlation under ideal conditions [1]. Reverse correlation just takes the average of the stimulus vectors that directly precede a spike from the neuron of interest. This average is called the spike-triggered average (STA), which can be used as the weight vector.
Classical nonlinear systems analysis [3][4] can be used to estimate the nonlinear function of the second step.
References
- ^ a b Simoncelli, E. P., Paninski, L., Pillow, J. & Swartz, O. in The Cognitive Neurosciences 3rd edn (ed. Gazzaniga, M.) 327-338 (MIT, 2004)
- ^ Chichilnisky, E. J., A simple white noise analysis of neuronal light responses. Network: Computation in Neural Systems 12:199-213. (2001)
- ^ Marmarelis & Marmerelis, 1978. Analysis of Physiological Systems: The White Noise Approach. London: Plenum Press.
- ^ Korenberg, Sakai, and Naka, 1989. Dissection of neuron network in the catfish inner retina. III. Interpretation of spike kernels. Journal Neurophysiology. 61:1110-1120.
 | This neuroscience article is a stub. You can help Wikipedia by expanding it. |