João Correia
ABSTRACT
A Genetic Programming approach for the improvement of the performance of classifier systems through the synthesis of new training instances is explored. Genetic Programming is used to exploit shortcomings of classifiers systems and generate misclassified instances. The proposed approach performs multiple parallel evolutionary runs to generate a large number of potentially misclassified samples. A supervisor module determines which of the generated images have been misclassified and which should be added to the training set. New classifiers are trained based on the original training set augmented by the selected evolved instances. The results attained while using face detection classifiers are presented and discussed. Overall they indicate that significant improvements are attained when using multiple evolutionary runs.
- W. Kienzle, G. H. Bakír, M. O. Franz, and B. Schölkopf. Face detection -- efficient and rank deficient. In L. K. Saul, Y. Weiss, and L. Bottou, editors, Advances in Neural Information Processing Systems 17, pages 673--680. MIT Press, Cambridge, MA, 2005.Google Scholar
- P. Machado and A. Cardoso. All the truth about NEvAr. Applied Intelligence, Special Issue on Creative Systems, 16(2):101--119, 2002. Google ScholarDigital Library
- P. Machado, J. Correia, and J. Romero. Improving face detection. In A. Moraglio, S. Silva, K. Krawiec, P. Machado, and C. Cotta, editors, Genetic Programming, volume 7244 of Lecture Notes in Computer Science, pages 73--84. Springer Berlin / Heidelberg, 2012. Google ScholarDigital Library
- H. Rowley, S. Baluja, and T. Kanade. Rotation invariant neural network-based face detection. In Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, June 1998.Google Scholar
- P. Viola and M. Jones. Rapid object detection using a boosted cascade of simple features. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages I--511--I--518 vol.1, Hawaii, 2001.Google Scholar
Index Terms
- Improving haar cascade classifiers through the synthesis of new training examples
Recommendations
Improving Text Classification Accuracy by Training Label Cleaning
In text classification (TC) and other tasks involving supervised learning, labelled data may be scarce or expensive to obtain. Semisupervised learning and active learning are two strategies whose aim is maximizing the effectiveness of the resulting ...
Bootstrapped training of event extraction classifiers
EACL '12: Proceedings of the 13th Conference of the European Chapter of the Association for Computational LinguisticsMost event extraction systems are trained with supervised learning and rely on a collection of annotated documents. Due to the domain-specificity of this task, event extraction systems must be retrained with new annotated data for each domain. In this ...
Co-training on authorship attribution with very fewlabeled examples: methods vs. views
SIGIR '14: Proceedings of the 37th international ACM SIGIR conference on Research & development in information retrievalAuthorship attribution (AA) aims to identify the authors of a set of documents. Traditional studies in this area often assume that there are a large set of labeled documents available for training. However, in the real life, it is hard or expensive to ...
Comments