MultiSpread

Research objective

---

There is a great interest in controlling spreading processes in multilayer networks. It has great potential to make a significant impact on tackling important societal challenges, such as containing the spread of diseases, stopping the spread of extremist behaviour, or inhibiting the spread of fake news.

Multilayer networks are inherently complex as they aim to represent complex systems. In this project, we focus on them as they enable us to encompass more detailed information about a complex system than a traditional single-layer network resulting in more realistic modelling scenarios. Each layer in a multilayer structure represents one type of relationship, e.g. one layer can be the physical contact layer and another online contact layer. Different layers can also denote different kinds of relations, e.g. friendship on one layer and family ties on another. Those layers have their own dynamics and interplay with each other, mutually influencing their dynamics. Due to this highly heterogeneous nature of multilayer networks (a form of representation of even more complex, real-world systems) through which the dynamical processes spread and their non-linear character, neither the network dynamics nor the dynamic processes over these structures can be analytically described. This poses severe constraints on the controllability of spreading processes. The control, in the context of multilayer networks, can be looked at through the lenses of control theory. It means that we must apply a set of inputs (control actions) to selected nodes and monitor how the network behaviour changes in response to these inputs. The feedback from that observation will allow us to check how far we are from the desired output and undertake corrective action if necessary. It still becomes an open problem as to when and how a multilayer network can be controlled. To address the issues mentioned above, there is an urgent need for extensive research into the multilayer networks' controllability and suitable control mechanisms that would direct the progress of spreading processes in the desired way.

The fundamental research in the area of controlling spreading processes, which is the focus of this project, has a wide range of applications; one of them is changing human behaviour. Pandemics, gender imbalance, financial market instability, and users' behaviour in online gaming, which are the main case studies to be developed within this project, represent only a few challenges that modern society faces. For the first time in human history, we can process big data about the various interactions and activities of millions of individuals that can be represented as a multilayer network. Not only we can analyse the dynamics of such structures but also the dynamics on those networks, including disease spread, opinion formation or fake news propagation. They represent an increasingly important resource in the process of understanding the behaviour of individuals, groups and whole communities. Yet, there is no coherent and comprehensive approach to analyse (i) how the spreading processes in those multilayer networks look like and (ii) how we can influence and (iii) control those processes. All three components are crucial to advance our understanding of continuously changing people's behaviour and propose mechanisms that help to change this behaviour if desired or necessary.

Spreading processes over networks and their evolution are usually analysed by building models of a process that happens over random, scale-free, and small-world models. However, these do not reflect the complex nature of real-world multilayer networks and related processes with high enough accuracy for effective control. Building on the previous related research, the proposed transformative route forward to overcome the limitations of existing techniques is, therefore, in developing novel high accuracy data-driven multilayer network simulation models with the ability to accurately model and take into account the inherent dynamics of spreading processes taking place in those networks as well as the dynamic behaviour induced by the applied, proposed control mechanisms.

The main drive of this project is to understand the mechanisms behind spreading processes in real multilayer networks and use this knowledge to influence and control them. Thus, the main goal is to build a robust and adaptive framework that will address the following objectives:
OB1. To develop and evaluate methods for influencing spreading processes in multilayer networks
OB2. To develop control mechanisms of the multilayer network structure
OB3. To develop and evaluate methods for spreading processes control in multilayer networks
OB4. To evaluate proposed solutions on real multilayer networks

Research tasks

---

1. Influencing spreading processes in multilayer networks
2. Controllability and control in multilayer networks
3. Control of spreading processes in multilayer networks
4. Data acquisition, use cases, and results exploitation

Outcomes