/home/precedings/apps/precedings/shared/storage/document/2007/262/npre2007262-1

[ ]

1st INCF Workshop on Large-scale Modeling of the Nervous System

December 1213, 2006

International Neuroinformatics Coordinating Facility Secretariat

Stockholm, Sweden

Authors

Mikael Djurfeldt and Anders Lansner

Scientific Organizer

Anders Lansner, KTH, Stockholm, Sweden

Workshop Participants

David Beeman, University of Colorado at Boulder, Boulder, CO, USA
Andrew Davison, UNIC, Gif-sur-Yvette, France
Markus Diesmann, RIKEN Brain Science Institute, Wako, Japan
Rodney Douglas, UNIETH, Zürich, Switzerland
Jens Eberhard, Ruprecht-Karls-University of Heidelberg, Heidelberg, Germany
Frederick Harris, University of Nevada, Reno, NV, USA
Michael Hines, Yale University, New Haven, CT, USA
Thomas Natschläger, Software Competence Center Hagenberg, Hagenberg, Austria
Charles Peck, IBM Research Division, Yorktown Heights, NY, USA
Shiro Usui, RIKEN Brain Science Institute, Wako, Japan
Gabriel Wittum, Ruprecht-Karls-University of Heidelberg, Heidelberg, Germany
Rapporteur: Mikael Djurfeldt, KTH, Stockholm, Sweden

Supported by the INCF Central Fund and the Swedish Foundation for Strategic Research

Nature Precedings : doi:10.1038/npre.2007.262.1 : Posted 27 Jun 2007

[ ]

Contents

Executive Summary

Introduction

Concepts

3.1 What do we mean by "model"?

3.2

Model complexity

3.3

Abstraction and level of description

3.4

Detailed versus abstract models

3.5

Realism

Directions in modeling of the nervous system - scientific needs

4.1

The role of the model in current neuroscience

4.2

Top-down and bottom-up approaches

4.3

Explicitness

4.4

Large-scale models

4.5 Simulation versus emulation

4.6 Upscaling

Software for large-scale simulations

5.1 Important properties of simulator software

5.2 Diversity of simulators

5.3 Software interoperability

5.4 Accuracy of simulation

5.5 Preprocessing and specification of large-scale network models

5.6 Declarative versus procedural model description

5.7 Postprocessing and visualization

Infrastructure

6.1 The role of the INCF with regard to large-scale modeling

6.2 Verification of simulator function

6.3 Model verification

6.4 Model reproducibility

6.5 Method development in computational neuroscience

6.6 A cyber infrastructure for computational neuroscience?

Nature Precedings : doi:10.1038/npre.2007.262.1 : Posted 27 Jun 2007

[ ]

1 Executive Summary

The goal of this workshop was to survey current demands, on-

going activities, and plans relating to development of tools for

scalable neural network simulation. Areas covered included

software components for preprocessing/model setup, as well

as for storage, analysis, and visualization of results. Partici-

pants discussed the need for coordinated action in the field

with regard to model, method and tool development.

In the following summary, we present three major findings and

seven recommendations developed by workshop participants.

Findings

The workshop participants reported:

· on difficulties in reproducing simulations from pub-

lished articles

· that the current diversity of simulators creates vigor

in the field and has benefits for the validation of

models

· on the importance of facilitating software interoper-

ability and re-use of simulation software components

Recommendations

The following recommendations were made by the workshop

participants:

· Arrange an annual workshop on large-scale model-

ing with the aim of developing standards for improv-

ing reusability of software. A workshop in the same

format as the Telluride Workshops on neuromorphic en-

gineering would give an opportunity to teach large-scale

simulation technology while simultaneously presenting

an optimal environment for practitioners and students

to discuss, test, and evaluate approaches to software

interoperability.

· Implement a standard simulator test suite. 1. INCF

should promote development of a set of simulator

"benchmarks" for the purposes of verifying correctness

of computation and serving as a standard for simulator

performance. These should preferably be selected with

respect to published models not contrived by the simula-

tor developers. . INCF should coordinate and contrib-

ute to raising funding for the implementation of the test

suite by members of the labs that have developed the

simulators. . INCF should maintain web pages with the

results of the test suite for each simulator. These pages

should be updateable so that old information can be

superseded by current best practices in each simulator.

· Implement an experimental framework for connect-

ing software components. A feasibility study should be

performed regarding the possibility of on-line communi-

cation between different software modules, for example

two parallel simulators. INCF should allocate resources

for implementing a software library with the communi-

cation interface.

· Develop publication guidelines for ensuring the

reproducibility of simulations. INCF should develop

guidelines for publications with computer simulations

and publish these on the INCF homepage as a reference

for authors. INCF should also encourage publishers

to implement policies for making models available in

conjunction with publication.

· Encourage, support and fund work on method de-

velopment within computational neuroscience. INCF

should inform funding agencies about the need for

research on methods in computational neuroscience, and

on methods for large-scale simulation in particular.

· Encourage, support and fund the production of pub-

lications on concepts and techniques for large-scale

simulations. INCF should inform publishers about the

need for published work on methods within the field of

large-scale simulations and computational neuroscience

in general, and about the current growth in this area.

· Continue work on defining areas in need of support

and coordination. Workshop participants agreed that

the 1st INCF Workshop on Large-scale Modeling of the

Nervous System was successful. However, there is still

a great need for further work on coordination within the

field. INCF should therefore arrange further workshops

with the aim of defining areas in need of support and

coordination.

Nature Precedings : doi:10.1038/npre.2007.262.1 : Posted 27 Jun 2007

[ ]

2 Introduction

Computer simulation is an increasingly important tool in neu-

roscience research. As computational modeling techniques

become integrated with experimental neuroscience, more

knowledge can be extracted from existing experimental data.

Quantitative models help to explain experimental observations

and seemingly unrelated phenomena. They assist in generating

experimentally testable hypotheses and in selecting informa-

tive experiments.

Supercomputers are becoming ever-more powerful. It has even

become possible to simulate models of a scale corresponding to

substantial parts of a small mammal's neocortex with biologi-

cally detailed compartmental neuron models. At the same time,

recent progress in experimental techniques holds the promise

of supplying modelers with data of an unprecedented level of

detail. Large-scale models are becoming an essential tool in

bridging multiple levels of organization in the description and

understanding of the nervous system. However, large-scale

modeling brings many new challenges, and there is a sense in

the community that we could benefit from coordination efforts

in areas such as standardization, interoperability and verifica-

tion.

The purpose of this report is to summarize discussions and rec-

ommendations of the workshop and to answer the question of

how INCF can support the large-scale modeling community.

To this end, section surveys the current status of large-scale

modeling in neuroscience and the scientific needs. Section 5

reviews the status of current tools for large-scale modeling and

discusses topics in need of development with regard to these

tools. Finally, section discusses how INCF can support the

community directly or contribute to changes with regard to

other organizations and scientific infrastructure.

3 Concepts

In the following, some basic concepts with regard to modeling

and simulation of the nervous system will be restated and clari-

fied. This will provide a basis for a discussion of the various

current developments in the field.

3.1 What do we mean by "model"?

Mathematical models are the language of science. According

to Wikipedia, a mathematical model is an abstract model ex-

pressed in mathematical language. Further:

An abstract model is a theoretical construct that represents

something, with a set of variables and a set of logical and

quantitative relationships between them. Models in this sense

are constructed to enable reasoning within an idealized logical

framework and are an important component of scientific theo-

ries. Idealized here means that the model may make explicit

assumptions that are known to be false (or incomplete) in some

detail. Such assumptions may be justified on the grounds that

they simplify the model while, at the same time, allowing the

production of acceptably accurate solutions.

It should be remembered that a mathematical model of reality

should always be regarded as idealized in the sense above. At

least this holds true for all types of model considered in this

report.

3.2 Model complexity

The golden standard with regard to model building is well cap-

tured by words often attributed to a certain famous physicist:

"Everything should be made as simple as possible, but not

simpler." Model complexity can be measured in many ways.

For this discussion, two measures are particularly important:

1. the number of model parameters, and . the number of state

variables in the model, which will here be denoted as model

dimension.

A model is always tailored to answer a specific set of ques-

tions. For a physicist, the ideal is to pick the model with the

smallest number of parameters which will still be sufficient to

answer the scientific questions posed. This has several good

consequences:

1. A simple model can be analyzed and understood. More

complex aspects of nature can be understood through

the strategy of divide-and-conquer.

2. A simpler model expresses a simpler scientific hypoth-

esis in the sense of Occam's razor. It cuts away elements

that are irrelevant to the problem studied.

Nature Precedings : doi:10.1038/npre.2007.262.1 : Posted 27 Jun 2007

[ ]

. A model with fewer parameters is better constrained,

making it easier to falsify. A model with many param-

eters can easily be adapted to the results of any experi-

ment.

Of the two dimensions of simplicity mentioned, the first one

is most important. A model which has few parameters is more

likely to express a well defined piece of scientific knowledge

in the form of a strong, i.e. falsifiable, hypothesis, while the

lack of this kind of simplicity threatens all three of the benefits

above. A large model dimension can make the model harder to

analyze and understand but there are many techniques avail-

able for handling this kind of complexity.

In addition, it is important to make a distinction between free

parameters--parameters which are tuned by the modeler or

software to achieve a certain model behavior--and parame-

ters constrained by experimental data. A model should have

few free parameters in order not to lose benefit 3 above. Also,

all model behavior which is used to tune parameters is trans-

formed from an "output" of the model to an "input", i.e., it can

no longer be claimed as a result, or prediction, of the model.

3.3 Abstraction and level of description

The tool used to achieve a simple model is abstraction. The

system is described at a certain level and elements which are

not believed to be important for answering the scientific ques-

tions asked are taken away.

Neuroscience spans many levels of description from molecules

to behavior. But what does level mean? Churchland and Se-

jnowski (199) discuss three categories of levels. The structure

of the nervous system has many different spatial scales with

substructures such as molecules, synapses, neurons, micro-

circuits, networks, regions and systems (list slightly modified

from Churchland). We call these levels of organization. With

some good will, behavior could be added at the top of this

hierarchy.

Marr (198) described levels along a different dimension in his

levels of analysis: 1. the computational level of abstract prob-

lem analysis, . the algorithmic level, specifying a formal pro-

cedure to perform the task so that a given input will yield the

correct output, . the level of physical implementation. Marr

argued that a higher-level question was largely independent of

the levels below and could be analyzed independently of the

lower level. However, it should be noted that Marr used, to a

large extent, neurobiological considerations to constrain and

inspire his computational theories and algorithms.

Churchlands third category, levels of processing, will not be

discussed here.

For a model of the primate primary visual cortex, V1, Marr's

computational level would correspond to what computations

are being performed in V1 and why. The algorithmic level

would correspond to how the information being processed on

the computational level is represented and how the computa-

tions are carried out, while the level of physical implementa-

tion describes the actual computational elements performing

these computations.

A typical large-scale network model, with single- or multicom-

partment units, thus belongs to the level of physical implemen-

tation, since it deals with neurons and synapses. But it can still

be inspired, and constrained, from the other two levels, and can

embody principles from the "higher" levels.

So far, we have discussed levels of organization and levels of

analysis. An abstract model leaves out aspects of the descrip-

tion of reality in order to achieve simplicity. Sometimes this

means leaving out elements from a lower level of organization,

but it can also mean describing something at a higher level of

analysis. A model of visual processing in terms of filter banks

and kernels is considered more abstract than a model of neu-

ronal populations in V1. Thus, we also talk about level of ab-

straction.

3.4 Detailed versus abstract models

A detailed model is often considered the opposite of an ab-

stract model. In this report, a detailed model is defined to mean

a model which spans several levels of organization. A model

which spans the levels from networks to behavior can thus si-

multaneously be more abstract and more detailed compared to

a model restricted to a single but lower level of organization.

A model of brain imaging data which incorporates networks

of simple units can be more detailed than a statistical model of

an ion channel. Yet, because it leaves out so much of the detail

beneath the level of networks, it can also be more abstract than

the latter model.

3.5 Realism

In order to say something about reality, and in order for a cor-

responding hypothesis to be falsifiable, a model needs to be

well rooted in empirical data, i.e. formulated in a way that is

consistent with a large set of experimental data.

It should now be made clear that, just as a model can be for-

mulated on all of the levels of organization discussed in sec-

Nature Precedings : doi:10.1038/npre.2007.262.1 : Posted 27 Jun 2007

[ 8 ]

tion ., empirical data can be obtained on different levels of

organization. It should also be pointed out that models can be

formulated on higher levels of analysis (in the sense of sec-

tion .): A cognitive psychologist can retrieve behavioral data

and construct models at the algorithmic level, without a direct

reference to how these processes are physically implemented

in the brain. Brain imaging retrieves data at the levels of net-

works, regions and systems. Electrophysiology collects data at

many levels of organization and this data can be used to con-

struct models at the level of physical implementation.

A model based on data from a lower level of organization, or

formulated in terms of elements from multiple lower levels of

organization, is not necessarily more realistic than data formu-

lated and rooted at a higher level. Rather, a model is realistic

if it is well rooted in empirical data at the given level, if its pa-

rameters are well constrained by these data, and if it correctly

predicts data which has not been used to tune the model.

Not only the parameters of a model, but also the equations,

represent assumptions about reality. Some models are based on

first principles, that is, their equations are established laws of

physics. Models based on first principles, or models with equa-

tions which can be derived from the laws of physics, can be

trusted to a larger extent than other models, because they rely

less than other models on assumptions in the form of peculiari-

ties of the model equations or fitting of parameters.

4 Directions in modeling of

the nervous system - scientific

needs

4.1 The role of the model in current neuroscience

This section contains a brief discussion of the various roles

that models currently have in neuroscience. As stated above,

models are the language of science. We use models to

· formulate hypotheses regarding the function of the ner-

vous system

The activity of formulating a hypothesis in terms of a model

requires collection of experimental data. It is often discovered

that crucial data are missing. In this respect a model could be

considered

· a tool to identify what we don't know

Often, hidden contradictions and inconsistencies are revealed

during the formulation process, and it happens that models

don't yield expected results so that a further role of the model

· validating self-consistency of the description of a phe-

nomenon or function

If the confidence in the model is strong but the predictions dif-

fer from experiment the model can be used to

· falsify hypotheses

If phenomena in the model are unexpected or unobserved a

model can

· suggest new experiments

Finally, a model can be used as a

· platform for integrating knowledge

unifying experimental data from many sources in a consistent

manner.

4.2 Top-down and bottom-up approaches

The top-down approach to modeling most often means using

hypotheses on the computational or algorithmic levels as a

starting point when approaching the formulation of a model at

the level of physical implementation. These functional hypoth-

eses guide the formulation of the model at the implementation-

al level in terms of what elements are included in the model

and which experimental data are considered important. A func-

tional hypothesis can also complement experimental data in

the sense of giving additional constraints. For example, if we

Nature Precedings : doi:10.1038/npre.2007.262.1 : Posted 27 Jun 2007

[ 9 ]

have limited experimental data on the functional connectivity

between inhibitory and excitatory connectivity, an additional

functional constraint would be that the activity in the network

must not be allowed to grow in an uncontrolled manner.

A top-down approach can also start with an abstract model,

neglecting detail below the level of organization at which the

model is formulated. Succeeding models can then increase the

amount of detail, enabling them to account for a larger set of

experimental data. For example, a connectionist, rate-based

model can be developed into a spiking network model, fol-

lowed by a Hodgkin-Huxley style model. However, it should

be noted that models are more typically developed outside of

such sequences, at a specific level of abstraction and detail

suitable to the scientific questions posed.

The bottom-up approach, in contrast, means using the level of

physical implementation as a starting point with the hope of

capturing function. For example, what does the anatomy of the

cerebral cortex mean? If we can, from the physical level of

synapses, dendrites, neurons and networks, identify computa-

tional primitives of the cortex such primitives can be abstracted

and we can move up one level of analysis. This is one goal of

projects like DAISY (Kennedy, 00), FACETS (Meier, 00)

and Blue Brain (Markram and Peck, 00).

As with the top-down approach, a bottom-up approach can be

concerned with levels of organization. In this case, it means to

take a lower level of organization as a starting point for under-

standing the higher level.

In practise, the approach of a modeller is usually neither pure-

ly top-down nor purely bottom-up, as was already evident in

Marr's work. Also, in the bottom-up approach, the focus of

experiments and the choice of elements to include in the model

is largely guided by functional hypotheses.

4.3 Explicitness

We define model explicitness to mean the degree to which the

model is isomorphic with reality, or, how directly state vari-

ables of the model can be mapped to empirical data.

The degree of detail in a multi-compartment, Hodgkin-Huxley,

model of a neuron aids in making this type of model transpar-

ent in the sense above. During intracellular recording of a neu-

ron, it is often possible to block a subset of the ion channels in

order to directly measure the current of another channel subset,

or, in order to study the effect on cell behavior. The explicit-

ness of the Hodgkin-Huxley formalism then makes it easy to

perform a corresponding manipulation in the model. It is also

easy to identify and display individual currents in the model.

In comparison, an integrate-and-fire model is less transparent.

While recent versions (e.g. Brette and Gerstner) of this type of

model can faithfully reproduce a spike train, some state vari-

ables correspond to the phenomenological effect of the coor-

dinated action of multiple channel types in the real neuron. In

this case, it is easy to compare the spike trains, but it is not

as easy to map the dynamics of the model to the individual

currents of the real neuron. On the other hand, the simplicity

of the integrate-and-fire model makes it easier to analyze and

understand.

Another aspect of explicitness is that it supports the role of

the model as a platform for integrating knowledge. A transpar-

ent model is more likely to connect well to a wider range of

experimental data, even data which were not targeted when

constructing the model.

Workshop participants commented that, in the past, there have

been many sacrifices in explicitness in order to achieve per-

formance. Early connectionist models used simple rate-coding

units, enabling the simulation of models of networks on the

personal computers of the 1980s, while Hodgkin-Huxley type

models could only be simulated one neuron at a time. Today,

we can simulate large networks of neurons with thousands of

compartments.

4.4 Large-scale models

For the purposes of this report, a large-scale model is a model

with a high dimension, i.e. a model with a large number of

state variables (on the order of hundreds of millions or more).

Thus, a detailed model of one cortical column can be large-

scale while an abstract model of a large network encompassing

multiple columns is not necessarily large-scale.

4.4.1 Integrate-and-fire models

The classical integrate-and-fire model (MacGregor and Oliver,

19; Tuckwell, 1988) has one state variable per neuron, rep-

resenting the membrane potential. It is basically a linear leaky

integrator with a voltage threshold and a reset mechanism. The

main advantage of this type of model compared to Hodgkin-

Huxley type models is its simplicity. For example, it has far

fewer parameters, while it still captures essential features of

the neurons. Because it has fewer parameters it is easier to

adapt this type of model to experimental data. In this sense, it

is easier to achieve a certain level of realism with an integrate-

and-fire model than with a Hodgkin-Huxley type model, while,

with more effort and more data, the latter model can reach an

Nature Precedings : doi:10.1038/npre.2007.262.1 : Posted 27 Jun 2007

[ 10 ]

even higher degree of realism. Its simplicity also makes it pos-

sible to develop automated procedures for extracting param-

eters from data (Jolivet et al., 00; Keat et al., 001; Paninski

et al., 00). Because of the low dimension and mathematical

tractability, it is easier to analyze and understand this type of

model. Finally, the simulation of this type of models require

fewer computational resources making it possible to simulate

larger networks given the same hardware. For examples of

large integrate-and-fire-based network models, see Mehring

et al. (00); Aviel et al. (00); Tetzlaff et al. (00); Kumar

et al. (00); Morrison et al. (00a).

The integrate-and-fire paradigm has recently been developed

in three directions (Brette and Gerstner, 00):

· the addition of a quadratic or exponential term, yield-

ing a smooth spike initiation zone (Latham et al., 000;

Fourcaud-Trocme et al., 00);

· the addition of a second state variable, enabling model-

ing of subthreshold resonances or adaptation (Izhikev-

ich, 00; Richardson et al., 00);

· using active conductances to model synaptic inputs

(Destexhe et al., 00).

4.4.2 Hodgkin-Huxley models

A network model with multi-compartmental Hodgkin-Huxley

type units is more detailed than its integrate-and-fire counter-

part. It is more complex, both by having more parameters and

a larger model dimension. As has been discussed in section

..1, this requires more labor to determine parameters from

experimental data. Sometimes, not all data is available so that

parameters need to be determined more indirectly. The disad-

vantage is that this turns an output of the model into an input

(.). For example, if we tune the conductance of a K

chan-

nel in order to obtain the correct time course of an AHP, instead

of measuring this conductance, we can no longer claim that our

model predicts a correct AHP. In this case, however, the kind

of predictions which are of interest in a network model appear

at another level of organization within the model.

In section ., the explicitness of HH-type models was dis-

cussed. The higher level of detail allows this type of model

to be connected to a wider range of experimental data. The

presence of ionic currents allows for comparatively easy mod-

eling of pharmacological manipulations. The D extent of a

compartmental model allows for the synthesis of EEG and LFP

signals (Einevoll et al., 00).

4.4.3 The bottom-up approach to the cortical
column

Regardless of whether we use integrate-and-fire or Hodgkin-

Huxley type units in a network model, an important set of

parameters that currently largely lacks an experimental basis

is the set of connectivity parameters. Data from, for example,

Thomson et al. (00) and Binzegger et al. (00) gives the

statistics of connectivity between pairs of cell types. This type

of data has led to the use of random Gaussian (e.g. Brunel)

or random uniform (e.g. Haeusler and Maass) connectivity in

network models, consistent with such statistics. However, it is

reasonable to assume that the microcircuitry of a column has

more structure than that. Also, there is very limited data on the

structure of long-range connectivity.

The Blue Brain project (Markram and Peck, 00) aims to col-

lect data on individual cells, for example acquisition of cell

morphology through cell labeling and -photon microscopy,

and then using database techniques and specialized software to

reconstruct a virtual column. The superposition of reconstruct-

ed cells in D-space may give additional constraints needed to

get a more complete picture of microcircuitry. The aim is also

to simulate a large-scale network model of a complete column

with multicompartmental Hodgkin-Huxley-type units without

reference to functional hypotheses about the network. Thus,

the approach is essentially hard-core bottom-up.

Workshop participants commented on the difficulty in experi-

mentally determining the existence of a synaptic contact be-

tween cortical neurons since axonal processes can pass close to

the dendritic processes of neurons without forming a synapse.

In essence, the only way to determine anatomically if there is

a contact is by looking at it with an electron microscope. It is

also possible to determine the existence of a connection elec-

trophysiologically by recording from pairs of neurons (e.g.,

Thomson et al.).

One particularly interesting development with regard to the ac-

quisition of connectivity data was discussed at the workshop.

Denk and Horstmann (00) have developed a method called

"serial block-face scanning electron microscopy" or SBFSEM.

A microtome is placed in the chamber of a scanning electron

microscope. The face of the tissue sample is scanned and 0

0 nm slices cut away, generating stacks of thousands of im-

ages from which a bulk D volume can be reconstructed. The

acquired data has enough resolution to trace thin axons and

identify synapses. This method holds the promise of geometri-

cally reconstructing an entire neocortical minicolumn and ex-

tracting its circuitry.

Nature Precedings : doi:10.1038/npre.2007.262.1 : Posted 27 Jun 2007

[ 11 ]

Workshop participants concluded that the acquisition of data

at multiple levels of organization leads to a new scale of proj-

ects--a development which parallels the study of the genome

and studies in particle physics. There will be new kinds of

challenges associated with industrial-scale data acquisition in

projects of thousands of man-years.

4.4.4 Combining the top-down and bottom-up
approaches

A survey of existing literature in computational neuroscience

shows that the pure bottom-up approach to understanding net-

work function is very rare. Part of the explanation is that miss-

ing empirical data need to be complemented with functional

hypotheses in order to make progress possible. But a top-down

approach may have importance in other ways than replacing

lacking knowledge. During the workshop, the question arose

whether a correctly implemented, detailed computer replica of

the cortical column would by itself say very much about net-

work function. Because a detailed model can be complex and

hard to analyze and understand, modellers tend to see func-

tional hypotheses and abstract models as a necessary comple-

ment in dissecting cortical function. It should still be noted that

a computer replica is much more accessible to experimentation

than the living tissue in the sense that any set of state variables

can be logged and an arbitrary set of variables can be simulta-

neously perturbed in a precise fashion.

An example of how a top-down approach can be combined

with the bottom-up approach is given by the model of Lun-

dqvist et al. (00) (see Djurfeldt et al. for a large-scale ver-

sion of the model). The model is mainly designed to target the

question: Is neocortical microarchitecture consistent with the

hypothesis of attractor memory network function? Here, most

parameters of the neuron models are determined from experi-

mental data in a bottom-up manner. However, the connectivity

parameters are determined by combining a long-range connec-

tivity structure required for attractor memory network function

with the currently existing empirical constraints on connectiv-

ity mentioned in section ...

Another aspect of this model, and most or all other network

models, is that the parameters of a neuron type are replicated

over the population of model neurons, with or without random

perturbation, in a crystal-like manner. This means that even if

a large-scale model of this type has a large model dimension, it

can still have a comparatively small number of parameters and,

thus, be simple in the important sense (c.f. section .).

4.4.5 Volume simulation

Until now, Hodgkin-Huxley type models have represented the

most basic level of organization at which we simulate neurons

and circuits, with the exception of hybrid models also includ-

ing biochemical processes inside the cell. Data from the SBF-

SEM method mentioned in section .. opens up the prospect

of a full D volume simulation of a cortical column. During

the workshop, Gabriel Wittum presented initial attempts in this

direction at the level of a single cell. In the Hodgkin-Huxley

approach, the neuron is modelled as an electrical circuit. Here,

instead, the D volume in which the neuron is embedded is de-

scribed in its entirety by partial differential equations (PDEs)

and simulated using the solver µG (Bastian et al., 199; Wit-

tum, 00).

A model of the D volume can be based on

first principles in

the sense discussed in section .. In this case, the model is

based on Maxwell's equations which describe the dynamics of

the electromagnetic field.

4.4.6 Simulation of growth processes

In order to fully understand the cortical architecture, it is neces-

sary to understand the development and growth processes from

which it results. Within the DAISY project, initial steps are

currently taken to simulate the migration of neuroblasts. This

adds a requirement on the simulator which is not yet fulfilled

by standard neuron simulators such as Neuron and Genesis:

there is a need to quantize space. Simulation tools are being de-

veloped within DAISY to meet this demand. The solver µG is

based on a communication layer, DDD (Dynamic Distributed

Data), which allows computational loads to migrate within a

parallel computer during simulation. This layer could also be a

suitable substrate for the simulation of growth processes.

4.5 Simulation versus emulation

Alan Turing used the term simulation in a very specialized

sense. The term referred to the simulation by a digital com-

puter of a subject discrete-state machine, defined by a set of

state transitions, inputs and outputs.

In computational neuroscience, the term usually refers to the

computation of the numerical approximation of a solution over

time to the equations of a mathematical model. When per-

formed on a digital computer, such computations are subject

to the limitations of the computer. For example, some quantum

mechanical processes can not be simulated on a digital com-

puter. However, such limitations do not seem to be of any rele-

vance to current computational models of the nervous system.

Nature Precedings : doi:10.1038/npre.2007.262.1 : Posted 27 Jun 2007

[ 1 ]

By an emulation, we refer to a model of the nervous system

in the shape of a physical realization, for example in terms of

an electronic circuit. The projects DAISY and FACETS both

implement VLSI chips emulating neural circuits.

During the workshop, participants commented that while most

simulations are slower than real-time (time simulated is shorter

than the wall-clock time required to compute the solution), it

is possible to construct an artefact which can emulate a neural

circuit in real-time. The important consequence is that the arte-

fact can interact with the environment in real time and react to

a continuous flow of events occurring asynchronously. In some

cases this is also achievable with a simulation on a digital com-

puter, as exemplified by the dynamic-clamp technique (Sharp

et al., 199) or the goal of the Blue Brain project to simulate a

cortical column on a super-computer in real time.

4.6 Upscaling

A model of the bulk D volume of neural tissue using PDEs

(section ..), if well-rooted in empirical data, can be consid-

ered more realistic than the Hodgkin-Huxley model. It is based

on first principles (sections 4.4.5, 3.5) while the Hodgkin-

Huxley model is partly based on simplifying assumptions and

curve fitting. This means that we can use the 3D volume model

to validate the multicompartmental Hodgkin-Huxley model.

Gabriel Wittum reported on simulations of a single cell where

ephaptic interactions could be observed between two of the

dendritic processes of the cell itself. Such a phenomenon

would not arise in the Hodgkin-Huxley model. It is also clear

that the surrounding interstitial fluid, neuropil, and dendritic

processes have substantial effects on the electric behavior of

the cell membrane, diverging from the Hodgkin-Huxley mod-

el. Clearly, however, the D volume simulation is not a good

replacement for the Hodgkin-Huxley model, because it is com-

putationally heavier. The question then arises what alternatives

we have to the HH model.

An important answer is given by the D model itself: There

exist mathematical techniques for deriving a model at a coarser

scale from a model at a finer scale. This methodology is called

upscaling (Eberhard et al., 00). Through upscaling tech-

niques it might be possible to derive a candidate model which

might serve as a better replacement for the Hodgkin-Huxley

model.

5 Software for large-scale

simulations

5.1 Important properties of simulator software

When judging software for large-scale simulation, there are

many criteria that need to be examined, some of which will be

mentioned in this section. Brette et al. (00) reviews existing

tools for simulation of networks of spiking neurons.

· Model types supported What neuronal/synaptic/plas-

ticity models can be simulated?

· Accuracy Does the simulator give correct results?

Recent work (Brette, 00; Rudolph and Destexhe,

00; Morrison et al., 00b) has presented methods for

determining spike times in a simulator more precisely,

and has shown that this can have effects on discharge

statistics and temporal precision in resolving synaptic

inputs. This is discussed further in section . below.

· Scalability In general, it is difficult to write efficient

parallel implementations. How does speedup (simula-

tion time divided by simulation time on one processor)

scale with the number of processors used? Ideally it

should grow linearly. How does simulation time scale as

a function of the size of the model?

· Documentation How good is the documentation?

· Support How quickly will the developers respond to

bug-reports or feature requests?

· License and availability of source code In a research

environment, it is an advantage to have the source code

for the simulator available, and to have permission to

modify it. This is guaranteed for all software covered by

the GPL license from the Free Software Foundation and

some related licenses.

· Adaptability How easy is it to adapt the simulator to

your purpose? How easy is to to add new mechanisms?

· Portability Does it run on my preferred platform?

· Interoperability How easy is it to collaborate with oth-

ers using a different simulator? This is discussed further

in section . below.

· Is there a graphical interface?

· What analysis/post-processing tools are available?

Nature Precedings : doi:10.1038/npre.2007.262.1 : Posted 27 Jun 2007

[ 1 ]

5.2 Diversity of simulators

During the workshop, it was discussed whether it is sensible

for the community to split efforts into the large and growing

set of neuron simulators available today rather than focussing

on one or a few tools. However, it was also noted that there is

currently a strong ongoing development of simulation technol-

ogy, and that simulators tend to have unique strong points not

shared by others.

The workshop also identified the reproducibility of models as

a major problem (see section .). The diversity of simulators

might allow for simulating the model on a different platform

from that on which it was originally developed, thereby verify-

ing the reproducibility of results. This leads to the first major

finding:

Workshop participants agreed that the current diversity of

simulators creates vigor in the field and has benefits for the

validation of models.

5.3 Software interoperability

Given the diversity of simulators, it is important to find ways

to share the gains produced by the efforts, of both developers

and users, put into individual simulators. During the workshop

some approaches to simulator-independent modeling environ-

ments were discussed, which allow a model to be simulated

using more than one simulation engine. This is important for

verification of simulator accuracy, for the reproduction, test-

ing and extension of published models, and for collaboration

between modellers using different simulation tools. The ap-

proaches discussed were graphical environments, declarative

model specifications using XML, and procedural model speci-

fications using an API implemented in the Python program-

ming language (see sections .. and .). Another conclu-

sion was that more effort should be put into the development of

such simulator-independent environments (`meta-simulators').

Thus, the second major finding of the workshop was:

Workshop participants agree upon the importance of fa-

cilitating software interoperability and re-use of simulation

software components.

5.3.1 Modularity

The practise of dividing software into modules with well-de-

fined roles has many advantages. It eases development and

Michael Hines noted that "The reason why we keep rein-

venting the wheel is that we haven't got it quite round yet."

increases maintainability of the code. If such modules have

well-defined interfaces, modules can be re-used in other cir-

cumstances. Such an interface can be in the form of

1. an application programming interface (API), enabling

a module in the form of a compilation unit or a library

to be linked into an application. This includes the defini-

tion of data structures required to pass information

through the interface.

. a communication interface, enabling modules in the

form of processes to communicate while running simul-

taneously on the same or on different machines

. a file format, allowing the output of a module in the

form of an application to be read as input to another

application. Here, "input" can be a model specification.

In this case, the interface takes the shape of a model

specification language.

As an example of the first form of modularity, a simulator can

be divided into a simulator kernel, responsible for the dis-

tribution and allocation of data structures over a cluster, for

building the model on the nodes, and for performing the com-

putations during a simulation, and other modules required for

module specification, input and output. The simulator kernel

can be further divided into a solver, with the sole responsibility

of performing computations, and modules required for alloca-

tion, distribution etc.

Workshop participants discussed the possibility of on-line in-

teraction between simulators (see section ..). This would

entail modularity of the second type.

An example of the third type of modularity is neuroConstruct

which is a software application for creating D models of

networks of biologically realistic neurons through a graphi-

cal user interface (GUI) (.1.). neuroConstruct can import

morphology files in Genesis, Neuron, Neurolucida, SWC and

MorphML formats for inclusion in network models and can

generate model specification files for Genesis and Neuron. Ef-

forts put into developing neuroConstruct further will thus ben-

efit both the Genesis and Neuron communities. Note, though,

that the choice of two output formats is forced by the current

lack of a standard format for model description. This will be

discussed further in section ...

David Beeman presented how the next generation of Gene-

sis, Genesis (..), aims to foster collaborative modeling

through a rich set of interfaces.

Nature Precedings : doi:10.1038/npre.2007.262.1 : Posted 27 Jun 2007