Methods in Neuronal Modeling - 2nd Edition: From Ions to Networks (Computational Neuroscience)

Rating:
Summary: Computational Modeling
Review: Great book for the theorist and experimentalist! I used the section on Epilepsy and the Neural Code for a grant I wrote. This book is a great reference and time spent reading it is very well rewarded. I bought the 1st & 2nd editions which are very different. Both editions are worth buying if one is involved with computer modeling, computation, mathematics, and plain old fashion recording neurophysiology.

Rating:
Summary: good backround
Review: How do you write a review by multiple authors who are authorities in the respective fields of each chapter? Carefully. Does the review need to contain the usual metaphors which when finally finished leave a prospective researcher with feelings of Oprah's book store. No. So let's start. The book contains general overall experimental setups on the "dry" sort, giving the broad over all design or outline of the direction of the model (read calculations) then includes the formulas at the end each chapter. Extremely nice for incorporation into MathCad or Matlab. (this is not as simple as it sounds, moving a high verbage book into something meaningful on a computer can't always be with as much style and grace as one wants). Most of the models however are, for want of a better discripter, "realistic".I guess this means they are not movies. Whether a model is done using ionic balanced equations or transfer functions seems somewhat academic to me as long as the information gained is useful. The level at which one can "view" these models certainly depends on the math and assets one can throw at the problem. As an example without to much "sweat" most of the cable equations, represented either as diffusion or electrical can be simulated far better in Matlabs PDE toolbox using Finite Element Methods. This allows the model to be viewed from the same point of view at different aspects. And why some kind of transfer functions cannot be made, and used in classical control theory, I'm working on now. Most of the equations using standard algebraic formulas work better in Mathcad using iterative range variables in extinction type modeling. This allows for the injection of pharmalogical testing data based on frequency manipulation, along digital signal processing lines. And running out of room, why not just take the data, use it to train a neural net and then use a Segueno fuzzy inference system to solve for the equations of state, which works very well for large systems (heart)such as the McKay-Glass non-linear types?