It is often supposed that messages sent to the visual cortex by the...

this article. the mean light level and discards information about the absolute intensity in the image. The purpose of both The only effective remedy will be to study visual processing under conditions of natural stimulation. There operations would be to remove from the image behaviorhas been a general reluctance to use natural images or ally uninteresting aspects that are mostly dependent on movies in vision research, mostly due to the seemingly the conditions of illumination or the average structure intractable complexity of natural scenes, the need to of the environment, while preserving and emphasizing consider the animal's eye movements, and the obvious the differences between objects in the visual scene. One bias that results from choosing any one stimulus from can speculate further that each successive stage of the such a large set. On the other hand, given the large early visual system adapts to---and consequently disuncertainties about what actually happens during natu- cards---what appear to be constants in the neural repreral vision, studying the response to even one or a few sentation from the previous stage (Barlow, 1990).

We propose a biophysical mechanism for the high interspike interval variability observed in cortical spike trains. The key lies in the nonlinear dynamics of cortical spike generation, which are consistent with type I membranes where saddle-node dynamics underlie excitability (Rinzel & Ermentrout, 1989). We present a canonical model for type I membranes, the θ-neuron. The θ-neuron is a phase model whose dynamics reflect salient features of type I membranes. This model generates spike trains with coefficient of variation (CV) above 0.6 when brought to firing by noisy inputs. This happens because the timing of spikes for a type I excitable cell is exquisitely sensitive to the amplitude of the suprathreshold stimulus pulses. A noisy input current, giving random amplitude “kicks” to the cell, evokes highly irregular firing across a wide range of firing rates; an intrinsically oscillating cell gives regular spike trains. We corroborate the results with simulations of the Morris-Lecar (M-L) neural model with random synaptic inputs: type I M-L yields high CVs. When this model is modified to have type II dynamics (periodicity arises via a Hopf bifurcation), however, it gives regular spike trains (CV below 0.3). Our results suggest that the high CV values such as those observed in cortical spike trains are an intrinsic characteristic of type I membranes driven to firing by “random” inputs. In contrast, neural oscillators or neurons exhibiting type II excitability should produce regular spike trains.

This memory is embedded in the distribution of channel states in the spike initiation site. The nature and resolution of this memory depend on the size of the channel pool and on the kinetics and number of states of the channels. We hypothesize that the number of channels in the spike initiation zone may be optimized in some sense to give the reliability and accuracy discussed above, together with a short-term memory of the neuron's activity. In this context, it is interesting to mention the work of Marder, Abbott, Turrigiano, Liu, and Golowasch (1996) and Abbott et al. (1996), which demonstrates activity-dependent long-term changes in the properties of intrinsic membrane currents.

You might find this additional information useful... This article cites 69 articles, 38 of which you can access free at:

Abstract: Many biological neural network models face the problem of scalability because of the limited computational power of today’s computers. Thus, it is difficult to assess the efficiency of these models to solve complex problems such as image processing. Here, we describe how this problem can be tackled using event-driven computation. Only the neurons that emit a discharge are processed and, as long as the average spike discharge rate is low, millions of neurons and billions of connections can be modeled. We describe the underlying computation and implementation of such a mechanism in SpikeNET, our neural network simulation package. The type of model one can build is not only biologically compliant, it is also computationally efficient as 400 000 synaptic weights can be propagated per second on a standard desktop computer. In addition, for large networks, we can set very small time steps (less than 0.01 ms) without significantly increasing the computation time. As an example, this method is applied to solve complex cognitive tasks such as face recognition in natural images.

The discrimination between two spike trains is a fundamental problem for both experimentalists and the nervous system itself. We introduce a measure for the distance between two spike trains. The distance has a time constant as a parameter. Depending on this parameter, the distance interpolates between a coincidence detector and a rate difference counter. The dependence of the distance on noise is studied with an integrate-and-fire model. For an intermediate range of the time constants, the distance depends linearly on the noise. This property can be used to determine the intrinsic noise of a neuron.

this article that the choice of a variability model has a major, nontrivial impact on the encoding properties of the neural population. The immense variability of individual response parameters, such as tuning widths or correlation coef#cients, has also been neglected in most previous work. Although these parameter variations are always found in empirical data, they were considered functionally insignificant, and hence theoretical studies have almost always assumed uniform parameters throughout the population. We will show here that this uniform case is unfavorable in the sense that the introduction of parameter variability improves the encoding performance

this article, we show that for a general class of spiking neuron models, which includes, in particular, the leaky integrate-and-#re model (Lapicque, 1907; Knight, 1972) as well as nonlinear spiking models, all three cases can occur if the input current is periodic, while aperiodic currents induce reproducible responses. In addition to numerical simulations, we put forth a theoretical explanation of this property for aperiodic currents that oscillate around threshold. The conditions required for our explanation are not ful#lled by the nonleaky integrate-and-#re model---also called perfect integrator---which is never reliable

A basic aspect of understanding the neural code of a neuron (or a neural system), is the ability to form a dictionary from the stimuli presented to the neuron (or the system) and the patterns of spikes that the neuron responds with. As neurons may respond unreliably to their stimuli, such a dictionary will be stochastic by nature. If the neuron responds to many dierent stimuli in a similar way (i.e. the number of stimulus features that the neuron 'cares about' is small), then the dictionary can be compressed, without a signi cant loss of its properties. Here we apply the agglomerative information bottleneck algorithm to study the properties of the dictionary (and neural code) of the identi ed H1 neuron in the y visual system. We nd that the neural code dictionaries of dierent ies are highly compressible, suggesting that a small number of features are the key components of the H1 neural code. We also compare the encoded features of the dierent ies and nd similar general structure, but dierences in the details. 1