Talk:Neuronal noise

From Scholarpedia
Jump to: navigation, search

    This article is good, but several topics deserve at list a bit more coverage. For example:

    1. Thermal noise: Can it be dismissed this easily? Is there a good reference on this point? 2. Ionic conductance noise: The explanation here it not that clear. Some readers may not realize that this section is being revisited in the section on modeling conductance noise. 3. Synaptic bombardment: As the author notes, this is believed to be a dominant source of noise in vivo. As such, it deserves more detailed treatment. 4. Chaos or noise: This section seems mis-filed (under "Sources of Noise") and too short. The author should consider expanding this section, "promoting" it in the outline, and placing it before "Modeling Neuronal Noise." 5. For the reader's easy comprehension, the section on multiplicative conductance noise should refer to the previous section on ionic conductance noise.

    There are a few typos 1. In "Neuronal Response Variability", change "Holden 1976}" to "Holden 1976)". 2. In "Dynamics versus Noise", insert a space after "(White et al. 1996)". Also, this reference does not exist in the bibliography.

    (Another reviewer) I think this is a very concise entry that covers all the major sources of neuronal noise that one may think of. The author did a good job with the survey, but I personally find this approach kind of dry, so my two suggestions are more along the lines of "let's make this more interesting", rather than pointing out any mistakes.

    (1) The history of trying to build a realistic model neuron that fires irregularly is interesting, and indicates that generating high variability with realistic elements is not trivial. In short, Softky and Koch showed that biophysical model neuron would typically fire regularly, because the numbers of spikes would be smoothed out. Either some kind of coincidence detection mechanism (their proposed solution) was needed, or the inputs had to be balanced (Shadlen and Newsome, 1996). Then it turned out that the balance was not quite enough: the spike trains have to be correlated to some degree, as seen by Zador and Stevens (1999, I think) and as proposed by Salinas (2002) on theoretical grounds.

    (2) The very last comment in the article is the most tantalizing one. I think the author could mention some of those efforts in trying to figure out whether the high cortical variability has any functional input, besides the possibility of stochastic resonance. Ma et al, in Alex Pouget's lab (2006 or 2007, in Nature Neuroscience), have suggested that Poisson noise is special because it allows encoded quantities to be combined using appropriate probabilistic rules. This result is not set in stone, but at least gives an interesting idea of what variability could be used for. Also, Basalyga (2006) suggested that neuronal noise might be useful for making a neural network more robust to synaptic changes (synaptic noise). Some ideas in any such directions would make this more exciting for the reader, I think.

    Other comments:

    I don't know what's the official format, but think it's annoying that there are two sections with references. I suggest putting all the cited materials under "References" and, if there is anything left that was not cited in the main text, then put it under "Further Reading".

    The Holden (1976) reference should go after Gerstner and Kistler, in "Further Reading".

    In "From Genetic Processes to Brain Rhythms", the Gerstner reference (last sentence) should be Gerstner and Kistler (2002).

    In "Synaptic Release Noise" the second sentence runs into the third. Something editing mistake happened there.

    In "Static Connectivity Noise", it should be "cells belonging to a given class".

    In "Modeling Synaptic Input and Stein's Model", in the last sentence, it should be "When the mean input is balanced...".

    In "References", for the Fourcaud-Trocme one, it should be "spike generation".

    In "Quadratic Integrate-and-Fire Model", the reference should be Fourcaud-Trocme (with capital T).

    In "Gain: Suprathreshold vs Subthreshold", should be "and its gain (slope...) modified by noise".

    In "Stochastic Resonance", it should be "will in fact induce an output".

    In "Dynamics versus noise", a space is needed after (White et al., 1996).

    Reviewer B: This is a concise but good review of the topic. I have the following suggestions to improve it:

    `Neural response variability': `cortical cells with highly irregular firing': add references? `vanishingly correlated' ISIs in pyramidal cells: has this been shown experimentally? What are the references?

    `From genetic processes (...)' ref. to Gerstner: do you mean Gerstner and Kistler's book?

    `Chaos or noise?' It could be a good idea to discuss the papers by van Vreeswijk and Sompolinsky (1996, 1998) in that section.

    Figure 1: in the caption, `nu'-> corresponding greek symbol.

    `Modeling synaptic input...' reference to Salinas and Sejnowski, fluctuation-driven regime: this terminology must have been introduced long before Salinas and Sejnowski's paper.

    `Diffusion limit...' Vsyn is missprinted.

    Reviewer C

    Reading over the comments above, I agree that some sections (e.g., synaptic bombardment) deserve expansion. I found the section on gain (suprathreshold vs. subthreshold) so brief, with no citations, that it was not useful. That section needs expansion and referencing.

    I believe the distinction between references and further reading works fine.

    The section on multiplicative conductance noise should probably cite new work in that area: Goldwyn and Shea-Brown 2011 PLoS Comput Biol AND/OR Goldwyn et al. 2011 PRE Orio and Soudry 2012 PLoS One These groups found better implementations of Fox-style algorithms

    Personal tools

    Focal areas