Preface

This textbook for advanced undergraduate and beginning graduate students provides a
systematic introduction into the fields of neuron modeling, neuronal dynamics, neural cod-
ing, and neural networks. It can be used as a text for introductory courses on Computa-
tional and Theoretical Neuroscience or as main text for a more focused course on Neural
Dynamics and Neural Modeling at the graduate level. The book is also a useful resource for
researchers and students who want to learn how different models of neurons and descrip-
tions of neural activity are related to each other.
All mathematical concepts are introduced the pedestrian way: step by step. All chapters
are richly illustrated by figures and worked examples. Each chapter closes with a short
summary and a series of mathematical Exercises. On the authors’ webpage Python source
code is provided for numerical simulations that illustrate the main ideas and models of the
chapter (http://lcn.epfl.ch/∼gerstner/NeuronalDynamics.html).
The book is organized into four parts with a total of 20 chapters. Part I provides a general
introduction to the foundations of computational neuroscience and its mathematical tools.
It covers classic material such as the Hodgkin–Huxley model, ion channels and dendrites,
or phase plane analysis of two-dimensional systems of differential equations. A special
focus is put on the firing threshold for the generation of action potentials, in the Hodgkin–
Huxley models, as well as in reduced two-dimensional neuron models such as the Morris–
Lecar model.
Part II focuses on simplified models for the dynamics of a single neuron. It covers
nonlinear integrate-and-fire models with and without adaptation, in particular the quadratic
and exponential integrate-and-fire model, as well as the Izhikevich model and adaptive
exponential integrate-and-fire model. The question of noise in the neural dynamics is posed
and two classic descriptions of noise are presented. First, stochasticity arising from random
spike arrival: this approach leads to a noise term in the differential equation of the voltage,
and can be formulated as a Langevin equation. Second, intrinsic stochasticity of neurons
leading to an “escape” across the firing threshold even when the neuron is in the sub-
threshold regime: this approach leads to the framework of a Generalized Linear Model
which is systematically introduced and discussed in applications of neuronal coding and
decoding. The relation between the neuron models of Part II and biological data is
highlighted and systematic parameter optimization algorithms are presented.

https://doi.org/10.1017/CBO9781107447615.001 Published online by Cambridge University Press

 x Preface

Part III takes the simplified models derived in Part II and builds networks out of these.
The collective properties of the network dynamics are described in terms of equations for
the population activity also called the population firing rate. The conditions under which
population activity can be described by a standard rate model are identified.
Part IV makes the link from dynamics to cognition. The population activity equations
are used for an analysis of famous paradigms of computational and cognitive neuroscience,
such as the neural activity during decision making or memory retrieval. In Part IV we also
sketch the theory of learning in relation to synaptic plasticity. The book closes with a
fascinating application of the principles of neuronal dynamics to help patients suffering
from Parkinson’s disease.
A small fraction of the text of the present book is based on Spiking Neuron Models
(Cambridge University Press) which was first published in 2002 and has been reprinted
several times since then. In the meantime, the field has changed and we felt that a simple
update of Spiking Neuron Models for a second edition would not be enough to give credit
to the developments that have occurred.
Scientifically, the scope of Spiking Neuron Models was limited in several respects. First,
it mainly focused on linear integrate-and-fire models, and mentioned their nonlinear coun-
terparts only in passing. In the present book, nonlinear integrate-and-fire models are treated
in a full chapter. Second, adaptation was neglected in the treatment 10 years ago – mainly
because population equations for adaptive neurons were not yet available. In the present
book, adaptive integrate-and-fire models are covered at length in a separate chapter and
the population activity equations for adaptive neurons are derived. Third, while the Spike
Response Model with escape noise has always contained all the features of a Generalized
Linear Model (GLM), by the year 2002 the theory of GLMs had not yet found its way into
the field of neuroscience and was therefore simply absent from the original book. Given
the phenomenal rise of GLMs in neuroscience, the theory of GLMs for fitting neuronal
data is given a prominent role in this book. Finally, during teaching we always felt the
need to show famous applications of the principles of neuronal dynamics, such as retrieval
of contents from associative memories or decision dynamics and the neuroscience of free
will. The present book covers these topics.
On a more general level, we felt that it would be useful to have a book that is, from the
beginning, designed as a textbook rather than a monograph. Therefore, the present book
makes the link to experimental data more visible, has more explanatory text, and, last but
not least, provides a series of exercises that have already been tested in the classroom over
several years.
We hope that this book will be useful for students and researchers alike.

Wulfram Gerstner, Werner Kistler, Richard Naud, Liam Paninski

https://doi.org/10.1017/CBO9781107447615.001 Published online by Cambridge University Press

 Preface xi

Advice to the reader

Each chapter starts with a specific question and gives first intuitive answers in the first
section. As the chapter proceeds, the material gets more advanced, and the presentation
becomes more technical. For a first reading of the book, it is possible to read only the first
section, or first two sections, of each chapter and just glance at the subsequent sections.
More specific advice depends on the background. For example, readers who are new to
the field of computational neuroscience are advised to spend enough time with the classic
material of Part I, before they move on to Parts II and IV. The expert reader may skip Part I
completely and start directly with Part II.
In Part III, the main ideas are exposed in Chapters 12 and 15, which present the founda-
tions for the rate models in Part IV. The more technical chapters and sections of Part III can
be skipped at a first reading, but are necessary for a thorough understanding of the current
developments in the field of computational neuroscience.
Part IV contains applications of neuronal dynamics to questions of cognition and can be
read in any arbitrary order.
Sections marked by an asterisk (∗) are mathematically more advanced and can be omit-
ted during a first reading of the book.

Acknowledgements
We would like to thank our students, visitors, exchange students, and postdocs who care-
fully read and commented on at least two chapters each, some many more: Dane Corneil,
Andrea De Antoni, Mortiz Deger, Mohammad Faraji, Nicolas Frémaux, Felipe Gerhard,
Laureline Logiaco, Skander Mensi, Alexandre Payeur, Christian Pozzorini, Kerstin
Preuschoff, Tilo Schwalger, Alex Seeholzer, Hesam Setareh, Carlos Stein, Tim Vogels,
Friedemann Zenke, Lorric Ziegler.
The writing of the text was a joint work of the four authors. Werner Kistler and Wulfram
Gerstner were the authors of Spiking Neuron Models from which several sections sur-
vived. Liam Paninski was mainly involved in writing Chapters 9–11 of the present book
and gave valuable input to other chapters of Part II. Richard Naud contributed to writing
Chapters 1–11 and 14 with a leading role in some of these, made valuable comments and
suggestions for all other chapters, and was responsible for all the figures. Wulfram Gerstner
wrote the first drafts of Parts III and IV and contributed text to all other chapters.

Wulfram Gerstner, Werner Kistler, Richard Naud, Liam Paninski

https://doi.org/10.1017/CBO9781107447615.001 Published online by Cambridge University Press

https://doi.org/10.1017/CBO9781107447615.001 Published online by Cambridge University Press