Back to overview

CHIP:  Cross-scale High-res Investigation of Power-laws in neural networks

English title CHIP: Cross-scale High-res Investigation of Power-laws in neural networks
Applicant Herrmann Hans Jürgen
Number 162190
Funding scheme South Korea
Research institution Institut für Neuroinformatik Universität Zürich Irchel und ETH Zürich
Institution of higher education ETH Zurich - ETHZ
Main discipline Other disciplines of Physics
Start/End 01.03.2016 - 28.02.2019
Approved amount 245'783.00
Show all

Keywords (6)

criticality; cross-scale structural analysis ; network topology vs. function; power-law statistics deviations; network analysis of neuronal cultures ; neuronal cultures on MEA

Lay Summary (German)

CHIP: Cross-scale High-res Investigation of Power-laws in neural networks,Yoonkey Nam and Ruedi Stoop Lab
Lay summary

Eine der wichtigsten Herausforderungen dieser Tage ist es, zu verstehen, wie biologische neuronale Netzwerke wirklich funktionieren. Ungeachtet der grossen Fortschritte, welche in diesem Feld bereits erzielt wurden, sind wir immer noch vom grossen Ziel, die Arbeitsweise des menschlichen Hirns zu verstehen, weit entfernt. Das kommt daher, dass wir noch gelernt haben, diese Systeme vom richtigen Gesichtspunkt aus zu untersuchen. Sich entwickelnde neuronale Netze sind verschiedenen strukturellen Veränderungen unterworfen, bevor sie ihren funktionellen Zweck erfüllen können. Diese Veränderungen, welche wohlverstanden mit dem Lernprozess, der in neuronalen Netzen später stattfinden kann, wenig gemeinsam haben, sind bisher noch nicht genügend genau beschrieben worden. In unserem Projekt definieren und messen wir geeignete Netzwerk-Charakteristiken, um den Entwicklungsprozess einer prägnanten mathematischen Beschreibung zuzuführen. Unsere Arbeit teilt sich in zwei sich ergänzende Herausforderungen auf: einem biotechnologischen Teil, welcher die Experimente und Daten bereitstellt, und einem computationellen und analytischen Teil, der auf die mathematische Charakterisierung des Entwicklungsprozesses abzielt. Nachdem wir die geeigneten Messprotokolle mit den geeigneten Observablen festgelegt haben, wenden wir uns dem eigentlichen Ziel unserer Untersuchung zu: Welches ist die Bedeutung der strukturellen Veränderungen entlang der Entwicklung der biologischen neuronalen Netze, gesehen im Lichte der biologischen Informationsverarbeitung? Die Beantwortung dieser Frage ist nicht nur von grosser wissenschaftlicher, sondern auch von grosser technologischer und medizinischer Bedeutung.

Direct link to Lay Summary Last update: 03.03.2016

Responsible applicant and co-applicants



Natural data structure extracted from neighborhood similarity graphs
LorimerTom (2019), Natural data structure extracted from neighborhood similarity graphs, in Chaos, Solitons and Fractals, 119, 326.
Clustering: How much bias do we need?
LorimerTom, HeldJenny (2017), Clustering: How much bias do we need?, in Phil. Trans. A, 375, 20160293.
Avalanche and edge-of-chaos criticality do not necessarily co-occur in neural networks
Kanders Karlis, Lorimer Tom, Stoop Ruedi (2017), Avalanche and edge-of-chaos criticality do not necessarily co-occur in neural networks, in Chaos , 27, 047408.
Emergent Complexity from Nonlinearity, in Physics, Engineering and the Life Sciences
Stoop Ruedi (ed.) (2017), Emergent Complexity from Nonlinearity, in Physics, Engineering and the Life Sciences, Springer, Berlin.
Neural avalanches at the edge-of-chaos?
Kanders Karlis, Stoop Ruedi (2017), Neural avalanches at the edge-of-chaos?, in Emergent Complexity from Nonlinearity, ComoSpringer, Berlin.
Power laws in neuronal culture activity from limited availability of a shared resource
Joo S., Nam Yoonkey, Stoop Ruedi, Berger Damian, Lorimer Tom (2017), Power laws in neuronal culture activity from limited availability of a shared resource, in Emergent Complexity from Nonlinearity, ComoSpringer Proceedings in Physics 191, Berlin.
Frequency sensitivity in mammalian hearing from a fundamental nonlinear physics model of the inner ear
KandersK., LorimerT., GomezF. (2017), Frequency sensitivity in mammalian hearing from a fundamental nonlinear physics model of the inner ear, in Scientific Reports, 7, 9931.


Group / person Country
Types of collaboration
Prof. L.A. Bunimovich, Georgia Institute of Technology United States of America (North America)
- in-depth/constructive exchanges on approaches, methods or results
Prof. Yoonkey Nam, Dept. Bio-and Brain Engineering, KAIST, Daejeon, Korea Korean Republic (South Korea) (Asia)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
- Research Infrastructure
- Exchange of personnel

Scientific events

Active participation

Title Type of contribution Title of article or contribution Date Place Persons involved
Zusammenarbeit Chip Individual talk 4 verschiedene Vorträge KAIST 05.12.2016 Kaist, Daejeon, Korean Republic (South Korea) Nam Yoonkey;
Nolta 2016 Talk given at a conference Whole Conference Session 27.11.2016 Yugawara, Japan Nam Yoonkey; Herrmann Hans Jürgen;


Title Date Place
NDES 2016 20.06.2016 Reykjavik, Iceland

Associated projects

Number Title Start Funding scheme
173688 Novel technologies from knowing how networks work-NDES 2017 01.06.2017 Scientific Conferences


One of the most challenging scientific tasks is to understand how biological neural networks work. Despite the huge progress currently made, we posit that we are still far away from being able to explain the final goal of this struggle, the brain. That is, because we still have not learnt how to properly look at these systems. This we propose to learn by working with relatively simple neural networks that develop on a chip, which gives us the opportunity to look at them in a space-time manner, hoping to understand how they finally become biological distributed information-processing systems. The latter topic is not only of great scientific, but obviously also of great technological and medical relevance. Developing neuronal structures undergo several structural changes before they become functional, i.e. apt to serve their designed purpose. These changes have not been adequately described so far at the level of the topology defining the network’s function. In our project, we measure the network characteristics during such a development and from these extract the most salient mathematical indicators of the process. This endeavor implies two challenges: A bio-technological one (i.e. how can the network characteristics be sampled adequately in space and time) and a mathematical one (what are the appropriate descriptors of these evolutions). Once we have defined the appropriate measurement protocols and identified the appropriate observables, there is a third challenge: What is the interpretation of these changes, seen from the angle of biological information processing? Obviously, that question can only be answered by the close collaboration of the two sides involved.On a practical level, we will address this challenge from growing neural networks of different origins (species, cell types,..) on a chip. To see the salient properties inherent to all of these networks, we will compare the time, and modus of the structural changes, also under conditions of externally added chemicals (e.g. growth factors,..) and under conditions of external stimulations. By means of external stimulations, we may even trigger or impede structural changes. Using this approach, we may even at some point, try to teach the system to control a device.While general technologies enabling the growth of neural networks on a chip have recently made huge progress, the mathematical description of the developing network structures still lags behind. A recent progress is the discovery of a preferential attachment paradigm that rules most of the biological networks that once were believed to be purely “random networks” in the terms of the mathematician Renyi. The consequence of this principle is the appearance of power-law distributions of network characteristics such as node degree, link weights, or size of the activated subnetwork in the case of stimulated neural networks.The study of complex networks has, in the recent past, pursued an understanding of macroscopic behavior by focusing on power-laws in microscopic observables; deviations from power-laws have been explained away by reference to the thermodynamic limit. We have developed a framework of classification of mesoscale structures (such as can be expected to exist and change during our biological network’s development) that we can apply directly to our problem. In our approach, to understand a class of prominent power-law deviations found ubiquitously in real-world networks, we introduce a network growth algorithm that extends preferential attachment with a novel edge saturation principle. With this approach we hope to establish a paradigm shift in the physics of complex networks, toward the use of power-law deviations to relate meso-scale structure with macroscopic behavior. The detailed analysis of the process responsible for the deviations characteristic for the considered class of systems will tell us from macroscopic (deviated power law distributions) of the observables how matters change on the mesoscale that is mostly responsible for the collective information processing.After a recent visit of the Swiss partner in Korea triggered by the Swiss SNF, we have already embarked on a collaboration that combines the generation of suitable data and its specific analysis. With this proposal, to make a big step forward in understanding the process of developing neural networks, we envisage putting this collaboration on a somewhat larger scale than momentarily possible (visits and exchange of students are necessary). The designed project hosts a huge potential in health and technology, and it is built on complimentary strengths, fueled by a wish to work together that goes beyond merely exploiting the opportunities offered by the present funding framework.