Marie Sklodowska-Curie Individual Fellowship

Funding Agency: European Commission.

Grant Scheme: Horizon 2020, Marie Sklodowska-Curie Actions Individual Fellowship.

Project Title: COEVOLFRAMEWORK -- Unified Framework for Analysis of Co-evolutionary Systems.

Experienced Researcher: Dr. Siang Yew Chong

Supervisor: Professor Peter Tino


Today's challenges are marked by more frequent and wide-spread episodes of social, economic, political and environmental crisis. Co-evolutionary systems offer a natural perspective and powerful tools to help us understand conditions that affect populations of agents whose behavior changes in response to their interaction outcomes in situations of strategic decision-making. Studying these complex co-evolutionary systems remains an open challenge as rich structures in the models are not taken into account. The overarching aim of this project is to fill this major research gap with a unified, principled framework to analyze complex co-evolutionary systems. At the core of our approach is the graph representation of interacting agent behaviors where problem structures are fully captured by complete orientations in the graph and associated co-evolutionary dynamics by sampling processes on the graph. This project combines complementary expertise of the Experienced Researcher (Dr. Chong) in large co-evolutionary systems and the Supervisor (Professor Tino, University of Birmingham) in complex, adaptive and dynamical systems. Its vision is that the framework provides foundation for new modelling tools benefiting policy-makers, regulators, and academics through better understanding and predictive quality of real-world strategic decision-making systems.


Research Publication 1: S. Y. Chong, P. Tino, J. He and X. Yao, ``A New Framework for Analysis of Coevolutionary Systems - Directed Graph Representation and Random Walks,'' Evolutionary Computation, Vol. 27, No. 2, pp. 195-228, 2019.

Summary: A new framework is introduced to represent and study coevolutionary systems in a principled and unified manner. Directed graph (digraph) representation captures the full underlying structure of coevolutionary problems, which are specified by the set of all alternative solutions and their pairwise preference relations. Coevolutionary search of these solutions through variation and selection processes guided by interaction outcomes is modelled as a specific type of Markov chains - random walks on digraphs (coevolutionary Markov chains). Through this framework, both qualitative and quantitative characterizations of coevolutionary systems can be made.

Research Publication 2: S. Y. Chong, P. Tino and J. He, ``Coevolutionary Systems and PageRank,'' Artificial Intelligence, Online 30 August 2019.

Summary: A deep connection between coevolutionary systems and PageRank is established through the framework introduced earlier. The study here addresses the issue that one needs prior qualitative knowledge of coevolutionary digraphs to be able to perform quantitative characterizations on coevolutionary systems. A principled approach to measure and rank performance (importance) of strategies (solutions) is made. Under PageRank formalism, strategies (vertices in the coevolutionary digraph) transfer part of their authories to those that dominate them at pairwise interactions level. Authorities indicate the performances of strategies. Crucially, PageRank authorities under suitable normalization are quantity-wise the same as long-term visitation probabilities associated with coevolutionary Markov chains. This allows us to obtain theoretical results and subsequently develop useful tools for quantitative analysis of coevolutionary systems.