Learning Automated Agents from Historical Game Data via Tensor Decomposition

Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 9021)

Abstract

War games and military war games, in general, are extensively played throughout the world to help train people and see the effects of policies. Currently, these games are played by humans at great expense and logistically require many people to be physically present. In this work, we describe how to automatically create agents from historical data to replace some of the human players. We discuss why game-theoretic approaches are inappropriate for this task and the benefits of learning such agents. We formulate a tensor decomposition formulation to this problem that is efficiently solvable in polynomial time. We discuss preliminary results on real world data and future directions.

Keywords

War gaming Modeling Tensor decomposition 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Brennan, R.: Protecting the Homeland: Insights from Army Wargames. No. 1490. Rand Corporation (2002)Google Scholar
  2. 2.
  3. 3.
    Davidson, I.: Knowledge Driven Clustering. In: International Joint Conference on Artificial Intelligence (IJCAI) (2009)Google Scholar
  4. 4.
    Davidson, I., Gilpin, S., Walker, P.: Adversarial Event Behavior and its Analysis. Journal of Knowledge Discovery and Data Mining (DMKD), November 2012Google Scholar
  5. 5.
    Lawson, C.L., Hanson, R.J.: Solving least squares problems. Classics in Appl. Math., No.15. SIAM, Philadelphia (1995)Google Scholar
  6. 6.
    Carroll, J.D., Chang, J.: Analysis of individual differences in multidimensional scaling via an N-way generalization of “Eckart-Young” decomposition. Psychometrika 35 (1970)Google Scholar
  7. 7.
    Kolda, T., Bader, B.: Tensor Decompositions and Applications. SIAM Review 51(3), 455–500 (2009)CrossRefMATHMathSciNetGoogle Scholar
  8. 8.
    Mitchell, T.: Machine Learning. McGraw Hill (1997)Google Scholar
  9. 9.
    Norvig, R., Russel, S.: Artificial Intelligence: A Modern Approach. Pearson (2002)Google Scholar

Copyright information

© Springer International Publishing Switzerland 2015

Authors and Affiliations

  1. 1.United States NavyBethesdaUSA
  2. 2.Department of Computer ScienceUniversity of California, DavisDavisUSA

Personalised recommendations