Skip to main content

To what extent can design principles for complex networks be derived from the study of error propagation phenomenon in smart and bio-inspired network structures?

Final Report Summary - EPP (To what extent can design principles for complex networks be derived from the study of error propagation phenomenon in smart and bio-inspired network structures?)

The EPP project has been concerned with the question of whether and to what extent design principles for complex networks can be derived from the study of error propagation phenomena in modern network structures found in biological, transportation and communication systems.

Five key objectives of the project were:
(a) to develop a semantic network and a simulation tool for error propagation using real-time agents and example networks;
(b) to perform empirical studies of example networks that enable to examine their robustness;
(c) to analyse the effect of real and generated errors not only on the global properties, but also on the local properties of the network;
(d) to develop and test new structures for complex systems less sensitive to errors and attacks than known structures and to measure their effects on network performance;
(e) to combine results from simulations and empirical observations in order to investigate how a system can be best (re)modelled to be less sensitive to errors and attacks.

These ambitious goals have only been partially met with the most significant original research contributions revolving around the proposal of a new generic diffusion–based algorithm that is applicable in a wide range of scenarios. The algorithm is based on social propagation mechanisms (i.e. push and pull) and embraces two most widely employed models: the independent cascade model and the linear threshold model. The selected strategies have been studied in the context of different network structures: random, small–world, and scale free. A range of diffusion properties, such as scope, speed, and duration have also been investigated. The influence of the size of the initial set of seeds as well as the selection strategy of the seeds on the diffusion process have been analysed. Finally, a systematic empirical assessment of how the proposed diffusion simulation strategies are associated with scope and speed of spread over a range of network structures and diffusion parameters has been performed.

So far, it has not been possible to develop a simulation tool using real-time agents. However, a part of the project was focused on the integration of these two paradigms to meet soft real-time constraints within a distributed middleware based on both multi-agent system and Real Time Specification for Java. From a technological point of view of such integration, the obtained results showed that this type of activity is possible. The proposed integrated approach offers several advantages. For example, it is less sensitive to time fluctuations in thread synchronization, it is FIPA compliant, requires relatively small changes in existing architectures and it may be integrated more easily in large Java-based distributed applications.

The project also resulted in preparation and publication of a special issue of the New Generation Computing journal on "Propagation Phenomenon in Complex Networks: Theory and Practice" as well as a comprehensive edited book on "Propagation Phenomena in Real World Networks” to be published by Springer in their Intelligent Systems Reference Library series. Both edited volumes have been meant to bring attention to this very important subject area and attracted an excellent selection of contributions providing a starting point and a reference for future research.