Aggregation Under Bias: Rényi Divergence Aggregation and Its Implementation via Machine Learning Markets

  • Amos J. Storkey
  • Zhanxing Zhu
  • Jinli Hu
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 9284)


Trading in information markets, such as machine learning markets, has been shown to be an effective approach for aggregating the beliefs of different agents. In a machine learning context, aggregation commonly uses forms of linear opinion pools, or logarithmic (log) opinion pools. It is interesting to relate information market aggregation to the machine learning setting.

In this paper we introduce a spectrum of compositional methods, Rényi divergence aggregators, that interpolate between log opinion pools and linear opinion pools. We show that these compositional methods are maximum entropy distributions for aggregating information from agents subject to individual biases, with the Rényi divergence parameter dependent on the bias. In the limit of no bias this reduces to the optimal limit of log opinion pools. We demonstrate this relationship practically on both simulated and real datasets.

We then return to information markets and show that Rényi divergence aggregators are directly implemented by machine learning markets with isoelastic utilities, and so can result from autonomous self interested decision making by individuals contributing different predictors. The risk averseness of the isoelastic utility directly relates to the Rényi divergence parameter, and hence encodes how much an agent believes (s)he may be subject to an individual bias that could affect the trading outcome: if an agent believes (s)he might be acting on significantly biased information, a more risk averse isoelastic utility is warranted.


Probabilistic model aggregation Rényi divergence Machine learning markets 


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Abernethy, J., Frongillo, R.: A collaborative mechanism for crowdsourcing prediction problems. In: Advances in Neural Information Processing Systems 24 (NIPS 2011) (2011)Google Scholar
  2. 2.
    Bache, K., Lichman, M.: UCI machine learning repository (2013).
  3. 3.
    Barbu, A., Lay, N.: An introduction to artificial prediction markets for classification (2011). arXiv:1102.1465v3
  4. 4.
    Breiman, L.: Bagging predictors. Machine Learning 24(2), 123–140 (1996)MathSciNetzbMATHGoogle Scholar
  5. 5.
    Chen, Y., Wortman Vaughan, J.: A new understanding of prediction markets via no-regret learning. In: Proceedings of the 11th ACM Conference on Electronic Commerce (2010)Google Scholar
  6. 6.
    Cole, R., Fleischer, L.: Fast-converging tatonnement algorithms for the market problem. Tech. rep., Dept. Computer Science. Dartmouth College (2007)Google Scholar
  7. 7.
    Dani, V., Madani, O., Pennock, D., Sanghai, S., Galebach, B.: An empirical comparison of algorithms for aggregating expert predictions. In: Proceedings of the Conference on Uncertainty in Artificial Intelligence (UAI) (2006)Google Scholar
  8. 8.
    Dietrich, F.: Bayesian group belief. Social Choice and Welfare 35(4), 595–626 (2010)CrossRefMathSciNetzbMATHGoogle Scholar
  9. 9.
    Dietterich, T.G.: Ensemble methods in machine learning. In: Kittler, J., Roli, F. (eds.) MCS 2000. LNCS, vol. 1857, pp. 1–5. Springer, Heidelberg (2000) CrossRefGoogle Scholar
  10. 10.
    Domingos, P.: Why does bagging work? a Bayesian account and its implications. In: Proceedings KDD (1997)Google Scholar
  11. 11.
    Everingham, M., et al.: The 2005 PASCAL visual object classes challenge. In: Quiñonero-Candela, J., Dagan, I., Magnini, B., d’Alché-Buc, F. (eds.) MLCW 2005. LNCS (LNAI), vol. 3944, pp. 117–176. Springer, Heidelberg (2006) Google Scholar
  12. 12.
    Garg, A., Jayram, T., Vaithyanathan, S., Zhu, H.: Generalized opinion pooling. AMAI (2004)Google Scholar
  13. 13.
    Goldbloom, A.: Data prediction competitions - far more than just a bit of fun. In: IEEE International Conference on Data Mining Workshops (2010)Google Scholar
  14. 14.
    Green, K.: The \({{{\$}}}\)1 million Netflix challenge. Technology Review (2006)Google Scholar
  15. 15.
    van Hateren, J.H., van der Schaaf, A.: Independent component filters of natural images compared with simple cells in primary visual cortex. Proceedings: Biological Sciences 265(1394), 359–366 (1998)Google Scholar
  16. 16.
    Heskes, T.: Selecting weighting factors in logarithmic opinion pools. In: Advances in Neural Information Processing Systems 10 (1998)Google Scholar
  17. 17.
    Kahn, J.M.: A generative Bayesian model for aggregating experts’ probabilities. In: Proceedings of the 20th Conference on Uncertainty in Artificial Intelligence, pp. 301–308. AUAI Press (2004)Google Scholar
  18. 18.
    Lay, N., Barbu, A.: Supervised aggregation of classifiers using artificial prediction markets. In: Proceedings of ICML (2010)Google Scholar
  19. 19.
    Maynard-Reid, P., Chajewska, U.: Aggregating learned probabilistic beliefs. In: Proceedings of the Seventeenth Conference on Uncertainty in Artificial Intelligence, pp. 354–361. Morgan Kaufmann Publishers Inc. (2001)Google Scholar
  20. 20.
    Ottaviani, M., Sørensen, P.: Aggregation of information and beliefs in prediction markets, fRU Working Papers (2007)Google Scholar
  21. 21.
    Pennock, D., Wellman, M.: Representing aggregate belief through the competitive equilibrium of a securities market. In: Proceedings of the Thirteenth Conference on Uncertainty in Artificial Intelligence, pp. 392–400 (1997)Google Scholar
  22. 22.
    Bell, R.M., Koren, Y., Volinsky, C.: All together now: A perspective on the NETFLIX PRIZE. Chance 24 (2010)Google Scholar
  23. 23.
    Rubinstein, M.: An aggregation theorem for securities markets. Journal of Financial Economics 1(3), 225–244 (1974)CrossRefGoogle Scholar
  24. 24.
    Rubinstein, M.: Securities market efficiency in an Arrow-Debreu economy. American Economic Review 65(5), 812–824 (1975)Google Scholar
  25. 25.
    Rubinstein, M.: The strong case for the generalised logarithmic utility model as the premier model of financial markets. Journal of Finance 31(2), 551–571 (1976)CrossRefGoogle Scholar
  26. 26.
    Storkey, A.: When training and test sets are different: Characterising learning transfer. In: Lawrence, C.S.S. (ed.) Dataset Shift in Machine Learning, chap. 1, pp. 3–28. MIT Press (2009).
  27. 27.
    Storkey, A.: Machine learning markets. In: Proceedings of Artificial Intelligence and Statistics, vol. 15. Journal of Machine Learning Research W&CP (2011).
  28. 28.
    Storkey, A., Millin, J., Geras, K.: Isoelastic agents and wealth updates in machine learning markets. In: Proceedings of ICML 2012 (2012)Google Scholar
  29. 29.
    West, M.: Bayesian aggregation. Journal of the Royal Statistical Society 147, 600–607 (1984)CrossRefzbMATHGoogle Scholar
  30. 30.
    Wolpert, D.H.: Stacked generalization. Neural Networks 5(2), 241–259 (1992).

Copyright information

© Springer International Publishing Switzerland 2015

Authors and Affiliations

  1. 1.Institute of Adaptive Neural Computation, School of InformaticsThe University of EdinburghEdinburghUK

Personalised recommendations