Advertisement

Procedures with Incomplete Information

  • Duccio Piovani
  • Jelena Grujić
  • Henrik J. JensenEmail author
Chapter

Abstract

As mentioned at the beginning of the chapter an obvious short coming concerning application to real situations of the forecasting procedure as described so far is that we make use of complete knowledge of the entire space of agents and their interactions. To test the strength of the results against incomplete information a first attempt has been introducing an error in the interaction matrix used for the mean field treatment. This represents the situation in which an observer would have to measure the interactions between agents and does so with an error. This is possibly the biggest problem one would have to overcome when trying to describe real systems. As we will see the forecasting method has proven itself to be quite robust, yielding similar results in both models even in the presence of non negligible errors.

Copyright information

© Springer Nature Switzerland AG 2019

Authors and Affiliations

  • Duccio Piovani
    • 1
  • Jelena Grujić
    • 2
    • 3
  • Henrik J. Jensen
    • 4
    Email author
  1. 1.Head of Data Sciencenam.RParisFrance
  2. 2.Department of Computer Science, Sciences and Bioengineering Sciences Artificial Intelligence LaboratoryVrije Universiteit BrusselBrusselsBelgium
  3. 3.Computer Science Department, Faculty of ScienceMachine Learning Group, Université Libre de BruxellesBrusselsBelgium
  4. 4.Department of Mathematics and Centre for Complexity ScienceImperial College LondonLondonUK

Personalised recommendations