DPG Phi
Verhandlungen
Verhandlungen
DPG

Regensburg 2019 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

SOE: Fachverband Physik sozio-ökonomischer Systeme

SOE 2: Computational Social Science and Data Science I

SOE 2.1: Talk

Monday, April 1, 2019, 09:30–10:00, H17

Machine intelligence for network science and evolutionary dynamics? — •Jan Nagler1 and Marc Timme21Frankfurt School — 2TU Dresden

Does machine learning truly matter in network science or other sectors of statistical physics? Many network scientists, in particular those with a strong background in statistical physics, remain sceptical. This may be because computer science, traditionally, is more focussed on performance than on getting insights, offering transparency, being general or finding a minimal model. In [Timme & Nagler, News and Views: Pattern of Propagation, Nature Physics, in print] we argue that if fundamental principles underlying network dynamics are identified prior to the employment of intransparent black box feature extraction, not only hard tasks can be solved but also valuable insights may be provided. But this requires to frame our mathematical predictions according to the conditions under which the natural and artificial networks around us reveal themselves. Thus, this requires to bridge different disciplines through collaborations with researchers of complementary expertise. This talk aims to spread this message. We will exemplify this for seemingly universally optimal strategies (Generous Zero Determinant Strategies) and seemingly unresolvable (Prisoner's dilemma) conflicts of networked actors in complex noisy environments.

100% | Mobile Layout | Deutsche Version | Contact/Imprint/Privacy
DPG-Physik > DPG-Verhandlungen > 2019 > Regensburg