Abstract
Budget constraints arise in many computer vision problems. Computational costs limit many automated recognition systems while crowdsourced systems are hindered by monetary costs. We leverage wide variability in image complexity and learn adaptive model selection policies. Our learnt policy maximizes performance under average budget constraints by selecting “cheap” models for low complexity instances and utilizing descriptive models only for complex ones. During training, we assume access to a set of models that utilize features of different costs and types. We consider a binary tree architecture where each leaf corresponds to a different model. Internal decision nodes adaptively guide model-selection process along paths on a tree. The learning problem can be posed as an empirical risk minimization over training data with a non-convex objective function. Using hinge loss surrogates we show that adaptive model selection reduces to a linear program thus realizing substantial computational efficiencies and guaranteed convergence properties.
Chapter PDF
Similar content being viewed by others
References
Anstreicher, K.M., Ji, J., Potra, F.A., Ye, Y.: Probabilistic analysis of an infeasible-interior-point algorithm for linear programming. Math. Oper. Res. 24(1), 176–192 (1999)
Bennett, K.P., Mangasarian, O.L.: Bilinear separation of two sets in n-space. Computational Optimization and Applications 2 (1993)
Bishop, C.M., et al.: Pattern recognition and machine learning, vol. 1. Springer, New York (2006)
Busa-Fekete, R., Benbouzid, D., Kégl, B.: Fast classification using sparse decision dags. In: 29th International Conference on Machine Learning (ICML) (2012)
Chen, M., Xu, Z., Weinberger, K.Q., Chapelle, O., Kedem, D.: Classifier cascade: Tradeoff between accuracy and feature evaluation cost. In: International Conference on Artificial Intelligence and Statistics (AISTATS), pp. 235–242 (2012)
Dalal, N., Triggs, B.: Histograms of oriented gradients for human detection. In: IEEE Computer Society Conference on Computer Vision and Pattern Recognition, CVPR 2005, vol. 1, pp. 886–893. IEEE (2005)
Dulac-Arnold, G., Denoyer, L., Preux, P., Gallinari, P.: Datum-wise classification: a sequential approach to sparsity. In: Gunopulos, D., Hofmann, T., Malerba, D., Vazirgiannis, M. (eds.) ECML PKDD 2011, Part I. LNCS, vol. 6911, pp. 375–390. Springer, Heidelberg (2011)
Gao, T., Koller, D.: Active classification based on value of classifier. In: NIPS, vol. 24, pp. 1062–1070 (2011)
He, H., Daume III, H., Eisner, J.: Imitation learning by coaching. In: Advances In Neural Information Processing Systems (NIPS), pp. 3158–3166 (2012)
Jiang, J., Teichert, A.R., Daumé III, H., Eisner, J.: Learned prioritization for trading off accuracy and speed. In: NIPS, pp. 1340–1348 (2012)
Karayev, S., Baumgartner, T., Fritz, M., Darrell, T.: Timely object recognition. In: NIPS, pp. 899–907 (2012)
Karayev, S., Fritz, M., Darrell, T.: Dynamic feature selection for classification on a budget. In: International Conference on Machine Learning (ICML): Workshop on Prediction with Sequential Models (2013)
Maaten, L., Welling, M., Saul, L.K.: Hidden-unit conditional random fields. In: International Conference on Artificial Intelligence and Statistics, pp. 479–488 (2011)
Megiddo, N.: On the complexity of polyhedral separability. Discrete & Computational Geometry 3(1) (1988)
Patterson, G., Hays, J.: Sun attribute database: Discovering, annotating, and recognizing scene attributes. In: Proceeding of the 25th Conference on Computer Vision and Pattern Recognition (CVPR) (2012)
Ross, S., Bagnell, D.: Efficient reductions for imitation learning. In: International Conference on Artificial Intelligence and Statistics (AISTATS), pp. 661–668 (2010)
Taskar, B., Guestrin, C., Koller, D.: Max-margin markov networks. In: NIPS (2003)
Trapeznikov, K., Saligrama, V.: Supervised sequential classification under budget constraints. In: International Conference on Artificial Intelligence and Statistics (AISTATS) (2013)
Tsochantaridis, I., Joachims, T., Hofmann, T., Altun, Y., Singer, Y.: Large margin methods for structured and interdependent output variables. Journal of Machine Learning Research 6(9) (2005)
Viola, P., Jones, M.: Robust Real-time Object Detection. International Journal of Computer Vision 4, 34–47 (2001)
Wang, J., Saligrama, V.: Local supervised learning through space partitioning. Advances in Neural Information Processing Systems 25 (2012)
Wang, J., Saligrama, V.: Locally-Linear Learning Machines (L3M). In: Asian Conference on Machine Learning, pp. 451–466 (2013)
Wang, J., Trapeznikov, K., Saligrama, V.: An LP for Sequential Learning Under Budgets. In: Proceedings of the Seventeenth International Conference on Artificial Intelligence and Statistics, pp. 987–995 (2014)
Weiss, D., Sapp, B., Taskar, B.: Dynamic structured model selection. In: ICCV (2013)
Xiao, J., Hays, J., Ehinger, K.A., Oliva, A., Torralba, A.: Sun database: Large-scale scene recognition from abbey to zoo. In: 2010 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 3485–3492. IEEE (2010)
Xu, Z., Kusner, M., Chen, M., Weinberger, K.Q.: Cost-sensitive tree of classifiers. In: Proceedings of the 30th International Conference on Machine Learning (ICML 2013), pp. 133–141 (2013)
Zhang, C., Zhang, Z.: A Survey of Recent Advances in Face Detection. Tech. rep., Microsoft Research (2010)
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
1 Electronic Supplementary Material
Rights and permissions
Copyright information
© 2014 Springer International Publishing Switzerland
About this paper
Cite this paper
Wang, J., Bolukbasi, T., Trapeznikov, K., Saligrama, V. (2014). Model Selection by Linear Programming. In: Fleet, D., Pajdla, T., Schiele, B., Tuytelaars, T. (eds) Computer Vision – ECCV 2014. ECCV 2014. Lecture Notes in Computer Science, vol 8690. Springer, Cham. https://doi.org/10.1007/978-3-319-10605-2_42
Download citation
DOI: https://doi.org/10.1007/978-3-319-10605-2_42
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-10604-5
Online ISBN: 978-3-319-10605-2
eBook Packages: Computer ScienceComputer Science (R0)