Network Structure and the Exploration and Exploitation of Complex Technological Landscapes

November 30, 2017

On November 13, the speaker at the CNS Research Colloquium was Tamer Khraisha, who gave a talk on the topic of "Network Structure and the Exploration and Exploitation of Complex Technological Landscapes". Tamer is a third year PhD student at the Center for Network Science with particular interests in Innovation Networks, Diffusion of Financial Innovations and Economic Networks.

The research project is rooted in the observation that technological innovation is nowadays increasingly stylized as a collective phenomenon of interactions between a multitude of innovating agents. The rationale behind the collective aspect of innovation is mainly attributed to the fact that information about known solutions can be dispersed among individual innovators, therefore making the development of a new innovation go beyond the inventive capacities of individuals. In this talk, Tamer demonstrated the results of an evolutionary agent-based model in which a group of firms collectively searched a complex (rugged) technological landscape and observed each other’s solutions with different frequencies through different observation networks.


Figure 1: A technological landscape can be illustrated as an actual landscape on which agents want to find the highest points representing the fittest solutions. In the model, nodes (which are firms) are placed randomly on the fitness landscape. At each iteration, each node will with some probability observe the solution of a random neighbor in the underlying network. If the fitness value of the solution of the neighbor is higher, the node will move to that solution with some probability. Otherwise, the node will explore the fitness landscape in isolation by choosing a random solution next to the current position.

As the main result, Tamer found that collective exploration improved average performance over independent exploration because good solutions could diffuse through the network at an early stage. Prior research on complex problem solving by collectives has found that the efficiency of networks, meaning the speed at which networks disseminate information, can have a relevant effect on the performance of the collective. Following up on these results, Tamer examined eight observation networks, which varied in terms of efficiency. He found that efficient networks outperformed inefficient networks, independently of the size of the network and the frequency of observation. In the end of his lecture Tamer also discussed results relating to the performance of more realistic networks and a case study from financial innovation, which makes his work the first model that considers realistic corporate behaviors and core-periphery structures.

Blog post by András Borsos

Attachment: 
Category: 

Share