Finding regions of uncertainty in learned models: An application to face detection

  • Shumeet Baluja
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 1498)


After training statistical models to classify sets of data into predetermined classes, it is often difficult to interpret what the models have learned. This paper presents a novel approach for finding examples which lie on the decision boundaries of statistical models trained for classification. These examples provide insight into what the model has learned. Additionally, they can provide candidates for use as additional training data for improving the performance of the statistical models. By labeling the examples which lie on the decision boundaries, we provide information to the model in the regions in which it is most uncertain. The approaches presented in this paper are demonstrated on the real-world vision-based task of detecting faces in cluttered scenes.


Face Detection Decision Boundary Scenery Image Standard Genetic Algorithm Lighting Correction 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    L. Atlas, D. Cohn, and R. Ladner. Training connectionist networks with queries and selective sampling. In Advances in Neural Information Processing Systems 2. Morgan Kaufmann, 1989.Google Scholar
  2. 2.
    Shumeet Baluja. Genetic algorithms and explicit search statistics. In M. Mozer, M. Jordan, and T. Petsche, editors, Advances in Neural Information Processing Systems (NIPS) 9. MIT Press, Cambridge, MA, 1998.Google Scholar
  3. 3.
    H. M. Cartwright and S. P. Harris. The application of the genetic algorithm to two-dimensional strings: The source apportionment problem. In Fifth International Conference on Genetic Algorithms. Morgan Kaufmann, 1989.Google Scholar
  4. 4.
    D. Cohn, Z. Ghahramani, and M.I. Jordan. Active learning with statistical models. In Advances in Neural Information Processing Systems 7. MIT Press, 1995.Google Scholar
  5. 5.
    Harris Drucker, Robert Schapire, and Patrice Simard. Boosting performance in neural networks. International Journal of Pattern Recognition and Artificial Intelligence, 7(4):705–719, 1993.CrossRefGoogle Scholar
  6. 6.
    John Hertz, Anders Krogh, and Richard G. Palmer. Introduction to the Theory of Neural Computation. Addison-Wesley, Reading, MA., 1991.Google Scholar
  7. 7.
    Henry A. Rowley, Shumeet Baluja, and Takeo Kanade. Neural network-based face detection. IEEE Transactions on Pattern Analysis and Machine Intelligence, 20:23–38, January 1998.CrossRefGoogle Scholar
  8. 8.
    Kah-Kay Sung. Learning and Example Selection for Object and Pattern Detection. PhD thesis, MIT AI Lab, January 1996. Available as AI Technical Report 1572.Google Scholar
  9. 9.
    D. Whitley and T. Starkweather. Genitor ii: A distributed genetic algorithm. JETAI, 2, 1990.Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 1998

Authors and Affiliations

  • Shumeet Baluja
    • 1
    • 2
  1. 1.Justsystem Pittsburgh Research CenterPittsburgh
  2. 2.School of Computer ScienceCarnegie Mellon UniversityPittsburgh

Personalised recommendations