May 24, 2019 | 15:00—16:00
Networked systems are ubiquitously present in the real world – food webs, inter-bank markets or communication networks, for instance. Prediction of collapses in networked systems is an extremely important, still almost impossible task. The main obstacle is typically the enormous size of the network resulting in extremely large amount of information necessary to describe its structure. Moreover, in many systems the network structure remains hidden. Based on the corollary of the famous Perron-Frobenius theorem called eigenvector quantization, we show that for a broad class of networked systems it is possible to detect the last stage before the crash without knowing the network structure and consequently to predict the collapse of the whole system.