[{"author":[{"first_name":"Anastasia","last_name":"Pentina","id":"42E87FC6-F248-11E8-B48F-1D18A9856A87","full_name":"Pentina, Anastasia"},{"full_name":"Sharmanska, Viktoriia","id":"2EA6D09E-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0003-0192-9308","last_name":"Sharmanska","first_name":"Viktoriia"},{"orcid":"0000-0001-8622-7887","first_name":"Christoph","last_name":"Lampert","id":"40C20FD2-F248-11E8-B48F-1D18A9856A87","full_name":"Lampert, Christoph"}],"language":[{"iso":"eng"}],"publist_id":"5243","conference":{"name":"CVPR: Computer Vision and Pattern Recognition","start_date":"2015-06-07","end_date":"2015-06-12","location":"Boston, MA, United States"},"publication_status":"published","month":"06","doi":"10.1109/CVPR.2015.7299188","fulldoi":"https://doi.org/10.1109/CVPR.2015.7299188","corr_author":"1","article_processing_charge":"No","arxiv":1,"oa":1,"publisher":"IEEE","date_updated":"2025-06-11T07:19:52Z","quality_controlled":"1","oa_version":"Preprint","external_id":{"arxiv":["1412.1353"]},"main_file_link":[{"url":"http://arxiv.org/abs/1412.1353","open_access":"1"}],"page":"5492 - 5500","year":"2015","abstract":[{"lang":"eng","text":"Sharing information between multiple tasks enables algorithms to achieve good generalization performance even from small amounts of training data. However, in a realistic scenario of multi-task learning not all tasks are equally related to each other, hence it could be advantageous to transfer information only between the most related tasks. In this work we propose an approach that processes multiple tasks in a sequence with sharing between subsequent tasks instead of solving all tasks jointly. Subsequently, we address the question of curriculum learning of tasks, i.e. finding the best order of tasks to be learned. Our approach is based on a generalization bound criterion for choosing the task order that optimizes the average expected classification performance over all tasks. Our experimental results show that learning multiple related tasks sequentially can be more effective than learning them jointly, the order in which tasks are being solved affects the overall performance, and that our model is able to automatically discover the favourable order of tasks. "}],"_id":"1857","date_published":"2015-06-01T00:00:00Z","department":[{"_id":"ChLa"}],"status":"public","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","title":"Curriculum learning of multiple tasks","day":"01","scopus_import":"1","citation":{"ista":"Pentina A, Sharmanska V, Lampert C. 2015. Curriculum learning of multiple tasks. CVPR: Computer Vision and Pattern Recognition, 5492–5500.","ama":"Pentina A, Sharmanska V, Lampert C. Curriculum learning of multiple tasks. In: IEEE; 2015:5492-5500. doi:<a href=\"https://doi.org/10.1109/CVPR.2015.7299188\">10.1109/CVPR.2015.7299188</a>","chicago":"Pentina, Anastasia, Viktoriia Sharmanska, and Christoph Lampert. “Curriculum Learning of Multiple Tasks,” 5492–5500. IEEE, 2015. <a href=\"https://doi.org/10.1109/CVPR.2015.7299188\">https://doi.org/10.1109/CVPR.2015.7299188</a>.","short":"A. Pentina, V. Sharmanska, C. Lampert, in:, IEEE, 2015, pp. 5492–5500.","ieee":"A. Pentina, V. Sharmanska, and C. Lampert, “Curriculum learning of multiple tasks,” presented at the CVPR: Computer Vision and Pattern Recognition, Boston, MA, United States, 2015, pp. 5492–5500.","apa":"Pentina, A., Sharmanska, V., &#38; Lampert, C. (2015). Curriculum learning of multiple tasks (pp. 5492–5500). Presented at the CVPR: Computer Vision and Pattern Recognition, Boston, MA, United States: IEEE. <a href=\"https://doi.org/10.1109/CVPR.2015.7299188\">https://doi.org/10.1109/CVPR.2015.7299188</a>","mla":"Pentina, Anastasia, et al. <i>Curriculum Learning of Multiple Tasks</i>. IEEE, 2015, pp. 5492–500, doi:<a href=\"https://doi.org/10.1109/CVPR.2015.7299188\">10.1109/CVPR.2015.7299188</a>."},"date_created":"2018-12-11T11:54:23Z","type":"conference"},{"type":"conference","citation":{"mla":"Lampert, Christoph. <i>Predicting the Future Behavior of a Time-Varying Probability Distribution</i>. IEEE, 2015, pp. 942–50, doi:<a href=\"https://doi.org/10.1109/CVPR.2015.7298696\">10.1109/CVPR.2015.7298696</a>.","ieee":"C. Lampert, “Predicting the future behavior of a time-varying probability distribution,” presented at the CVPR: Computer Vision and Pattern Recognition, Boston, MA, United States, 2015, pp. 942–950.","apa":"Lampert, C. (2015). Predicting the future behavior of a time-varying probability distribution (pp. 942–950). Presented at the CVPR: Computer Vision and Pattern Recognition, Boston, MA, United States: IEEE. <a href=\"https://doi.org/10.1109/CVPR.2015.7298696\">https://doi.org/10.1109/CVPR.2015.7298696</a>","short":"C. Lampert, in:, IEEE, 2015, pp. 942–950.","ista":"Lampert C. 2015. Predicting the future behavior of a time-varying probability distribution. CVPR: Computer Vision and Pattern Recognition, 942–950.","ama":"Lampert C. Predicting the future behavior of a time-varying probability distribution. In: IEEE; 2015:942-950. doi:<a href=\"https://doi.org/10.1109/CVPR.2015.7298696\">10.1109/CVPR.2015.7298696</a>","chicago":"Lampert, Christoph. “Predicting the Future Behavior of a Time-Varying Probability Distribution,” 942–50. IEEE, 2015. <a href=\"https://doi.org/10.1109/CVPR.2015.7298696\">https://doi.org/10.1109/CVPR.2015.7298696</a>."},"date_created":"2018-12-11T11:54:24Z","scopus_import":1,"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","title":"Predicting the future behavior of a time-varying probability distribution","day":"15","status":"public","date_published":"2015-10-15T00:00:00Z","department":[{"_id":"ChLa"}],"_id":"1858","abstract":[{"text":"We study the problem of predicting the future, though only in the probabilistic sense of estimating a future state of a time-varying probability distribution. This is not only an interesting academic problem, but solving this extrapolation problem also has many practical application, e.g. for training classifiers that have to operate under time-varying conditions. Our main contribution is a method for predicting the next step of the time-varying distribution from a given sequence of sample sets from earlier time steps. For this we rely on two recent machine learning techniques: embedding probability distributions into a reproducing kernel Hilbert space, and learning operators by vector-valued regression. We illustrate the working principles and the practical usefulness of our method by experiments on synthetic and real data. We also highlight an exemplary application: training a classifier in a domain adaptation setting without having access to examples from the test time distribution at training time.","lang":"eng"}],"year":"2015","main_file_link":[{"open_access":"1","url":"https://arxiv.org/abs/1406.5362"}],"external_id":{"arxiv":["1406.5362"]},"page":"942 - 950","date_updated":"2024-10-09T20:56:24Z","quality_controlled":"1","oa_version":"Preprint","publisher":"IEEE","fulldoi":"https://doi.org/10.1109/CVPR.2015.7298696","corr_author":"1","arxiv":1,"oa":1,"doi":"10.1109/CVPR.2015.7298696","publist_id":"5241","conference":{"end_date":"2015-06-12","start_date":"2015-06-07","name":"CVPR: Computer Vision and Pattern Recognition","location":"Boston, MA, United States"},"publication_status":"published","month":"10","author":[{"id":"40C20FD2-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0001-8622-7887","last_name":"Lampert","first_name":"Christoph","full_name":"Lampert, Christoph"}],"language":[{"iso":"eng"}]},{"month":"06","publication_status":"published","conference":{"end_date":"2015-06-12","start_date":"2015-06-07","name":"CVPR: Computer Vision and Pattern Recognition","location":"Boston, MA, USA"},"publist_id":"5240","ec_funded":1,"language":[{"iso":"eng"}],"project":[{"call_identifier":"FP7","_id":"2532554C-B435-11E9-9278-68D0E5697425","grant_number":"308036","name":"Lifelong Learning of Visual Scene Understanding"},{"name":"Discrete Optimization in Computer Vision: Theory and Practice","grant_number":"616160","call_identifier":"FP7","_id":"25FBA906-B435-11E9-9278-68D0E5697425"}],"author":[{"full_name":"Shah, Neel","id":"31ABAF80-F248-11E8-B48F-1D18A9856A87","last_name":"Shah","first_name":"Neel"},{"id":"3D50B0BA-F248-11E8-B48F-1D18A9856A87","first_name":"Vladimir","last_name":"Kolmogorov","full_name":"Kolmogorov, Vladimir"},{"full_name":"Lampert, Christoph","id":"40C20FD2-F248-11E8-B48F-1D18A9856A87","first_name":"Christoph","last_name":"Lampert","orcid":"0000-0001-8622-7887"}],"oa":1,"article_processing_charge":"No","arxiv":1,"fulldoi":"https://doi.org/10.1109/CVPR.2015.7298890","corr_author":"1","doi":"10.1109/CVPR.2015.7298890","oa_version":"Preprint","quality_controlled":"1","date_updated":"2025-06-11T07:20:12Z","publisher":"IEEE","year":"2015","external_id":{"arxiv":["1408.6804"]},"main_file_link":[{"url":"http://arxiv.org/abs/1408.6804","open_access":"1"}],"page":"2737 - 2745","_id":"1859","abstract":[{"lang":"eng","text":"Structural support vector machines (SSVMs) are amongst the best performing models for structured computer vision tasks, such as semantic image segmentation or human pose estimation. Training SSVMs, however, is computationally costly, because it requires repeated calls to a structured prediction subroutine (called \\emph{max-oracle}), which has to solve an optimization problem itself, e.g. a graph cut.\r\nIn this work, we introduce a new algorithm for SSVM training that is more efficient than earlier techniques when the max-oracle is computationally expensive, as it is frequently the case in computer vision tasks. The main idea is to (i) combine the recent stochastic Block-Coordinate Frank-Wolfe algorithm with efficient hyperplane caching, and (ii) use an automatic selection rule for deciding whether to call the exact max-oracle or to rely on an approximate one based on the cached hyperplanes.\r\nWe show experimentally that this strategy leads to faster convergence to the optimum with respect to the number of requires oracle calls, and that this translates into faster convergence with respect to the total runtime when the max-oracle is slow compared to the other steps of the algorithm. "}],"status":"public","department":[{"_id":"VlKo"},{"_id":"ChLa"}],"date_published":"2015-06-01T00:00:00Z","scopus_import":"1","day":"01","title":"A multi-plane block-coordinate Frank-Wolfe algorithm for training structural SVMs with a costly max-oracle","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","type":"conference","citation":{"ista":"Shah N, Kolmogorov V, Lampert C. 2015. A multi-plane block-coordinate Frank-Wolfe algorithm for training structural SVMs with a costly max-oracle. CVPR: Computer Vision and Pattern Recognition, 2737–2745.","ama":"Shah N, Kolmogorov V, Lampert C. A multi-plane block-coordinate Frank-Wolfe algorithm for training structural SVMs with a costly max-oracle. In: IEEE; 2015:2737-2745. doi:<a href=\"https://doi.org/10.1109/CVPR.2015.7298890\">10.1109/CVPR.2015.7298890</a>","chicago":"Shah, Neel, Vladimir Kolmogorov, and Christoph Lampert. “A Multi-Plane Block-Coordinate Frank-Wolfe Algorithm for Training Structural SVMs with a Costly Max-Oracle,” 2737–45. IEEE, 2015. <a href=\"https://doi.org/10.1109/CVPR.2015.7298890\">https://doi.org/10.1109/CVPR.2015.7298890</a>.","short":"N. Shah, V. Kolmogorov, C. Lampert, in:, IEEE, 2015, pp. 2737–2745.","ieee":"N. Shah, V. Kolmogorov, and C. Lampert, “A multi-plane block-coordinate Frank-Wolfe algorithm for training structural SVMs with a costly max-oracle,” presented at the CVPR: Computer Vision and Pattern Recognition, Boston, MA, USA, 2015, pp. 2737–2745.","apa":"Shah, N., Kolmogorov, V., &#38; Lampert, C. (2015). A multi-plane block-coordinate Frank-Wolfe algorithm for training structural SVMs with a costly max-oracle (pp. 2737–2745). Presented at the CVPR: Computer Vision and Pattern Recognition, Boston, MA, USA: IEEE. <a href=\"https://doi.org/10.1109/CVPR.2015.7298890\">https://doi.org/10.1109/CVPR.2015.7298890</a>","mla":"Shah, Neel, et al. <i>A Multi-Plane Block-Coordinate Frank-Wolfe Algorithm for Training Structural SVMs with a Costly Max-Oracle</i>. IEEE, 2015, pp. 2737–45, doi:<a href=\"https://doi.org/10.1109/CVPR.2015.7298890\">10.1109/CVPR.2015.7298890</a>."},"date_created":"2018-12-11T11:54:24Z"},{"main_file_link":[{"open_access":"1","url":"http://www.cv-foundation.org/openaccess/content_cvpr_2015/papers/Royer_Classifier_Adaptation_at_2015_CVPR_paper.pdf"}],"page":"1401 - 1409","date_created":"2018-12-11T11:54:24Z","citation":{"chicago":"Royer, Amélie, and Christoph Lampert. “Classifier Adaptation at Prediction Time,” 1401–9. IEEE, 2015. <a href=\"https://doi.org/10.1109/CVPR.2015.7298746\">https://doi.org/10.1109/CVPR.2015.7298746</a>.","ama":"Royer A, Lampert C. Classifier adaptation at prediction time. In: IEEE; 2015:1401-1409. doi:<a href=\"https://doi.org/10.1109/CVPR.2015.7298746\">10.1109/CVPR.2015.7298746</a>","ista":"Royer A, Lampert C. 2015. Classifier adaptation at prediction time. CVPR: Computer Vision and Pattern Recognition, 1401–1409.","apa":"Royer, A., &#38; Lampert, C. (2015). Classifier adaptation at prediction time (pp. 1401–1409). Presented at the CVPR: Computer Vision and Pattern Recognition, Boston, MA, United States: IEEE. <a href=\"https://doi.org/10.1109/CVPR.2015.7298746\">https://doi.org/10.1109/CVPR.2015.7298746</a>","short":"A. Royer, C. Lampert, in:, IEEE, 2015, pp. 1401–1409.","ieee":"A. Royer and C. Lampert, “Classifier adaptation at prediction time,” presented at the CVPR: Computer Vision and Pattern Recognition, Boston, MA, United States, 2015, pp. 1401–1409.","mla":"Royer, Amélie, and Christoph Lampert. <i>Classifier Adaptation at Prediction Time</i>. IEEE, 2015, pp. 1401–09, doi:<a href=\"https://doi.org/10.1109/CVPR.2015.7298746\">10.1109/CVPR.2015.7298746</a>."},"type":"conference","year":"2015","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","day":"01","publisher":"IEEE","title":"Classifier adaptation at prediction time","oa_version":"Submitted Version","quality_controlled":"1","date_updated":"2021-01-12T06:53:41Z","scopus_import":1,"doi":"10.1109/CVPR.2015.7298746","date_published":"2015-06-01T00:00:00Z","department":[{"_id":"ChLa"}],"fulldoi":"https://doi.org/10.1109/CVPR.2015.7298746","oa":1,"status":"public","_id":"1860","abstract":[{"lang":"eng","text":"Classifiers for object categorization are usually evaluated by their accuracy on a set of i.i.d. test examples. This provides us with an estimate of the expected error when applying the classifiers to a single new image. In real application, however, classifiers are rarely only used for a single image and then discarded. Instead, they are applied sequentially to many images, and these are typically not i.i.d. samples from a fixed data distribution, but they carry dependencies and their class distribution varies over time. In this work, we argue that the phenomenon of correlated data at prediction time is not a nuisance, but a blessing in disguise. We describe a probabilistic method for adapting classifiers at prediction time without having to retrain them. We also introduce a framework for creating realistically distributed image sequences, which offers a way to benchmark classifier adaptation methods, such as the one we propose. Experiments on the ILSVRC2010 and ILSVRC2012 datasets show that adapting object classification systems at prediction time can significantly reduce their error rate, even with no additional human feedback."}],"author":[{"full_name":"Royer, Amélie","last_name":"Royer","first_name":"Amélie"},{"first_name":"Christoph","last_name":"Lampert","orcid":"0000-0001-8622-7887","id":"40C20FD2-F248-11E8-B48F-1D18A9856A87","full_name":"Lampert, Christoph"}],"ec_funded":1,"language":[{"iso":"eng"}],"project":[{"call_identifier":"FP7","_id":"2532554C-B435-11E9-9278-68D0E5697425","grant_number":"308036","name":"Lifelong Learning of Visual Scene Understanding"}],"conference":{"start_date":"2015-06-07","name":"CVPR: Computer Vision and Pattern Recognition","end_date":"2015-06-12","location":"Boston, MA, United States"},"publist_id":"5239","month":"06","publication_status":"published"},{"article_processing_charge":"No","corr_author":"1","oa":1,"status":"public","date_published":"2014-05-10T00:00:00Z","department":[{"_id":"ChLa"}],"conference":{"location":"Beijing, China","end_date":"2014-06-26","name":"ICML: International Conference on Machine Learning","start_date":"2014-06-21"},"publist_id":"4844","intvolume":"        32","month":"05","publication_status":"published","abstract":[{"lang":"eng","text":"Transfer learning has received a lot of attention in the machine learning community over the last years, and several effective algorithms have been developed. However, relatively little is known about their theoretical properties, especially in the setting of lifelong learning, where the goal is to transfer information to tasks for which no data have been observed so far. In this work we study lifelong learning from a theoretical perspective. Our main result is a PAC-Bayesian generalization bound that offers a unified view on existing paradigms for transfer learning, such as the transfer of parameters or the transfer of low-dimensional representations. We also use the bound to derive two principled lifelong learning algorithms, and we show that these yield results comparable with existing methods."}],"_id":"2160","author":[{"id":"42E87FC6-F248-11E8-B48F-1D18A9856A87","first_name":"Anastasia","last_name":"Pentina","full_name":"Pentina, Anastasia"},{"full_name":"Lampert, Christoph","last_name":"Lampert","first_name":"Christoph","orcid":"0000-0001-8622-7887","id":"40C20FD2-F248-11E8-B48F-1D18A9856A87"}],"language":[{"iso":"eng"}],"type":"conference","year":"2014","main_file_link":[{"url":"https://dl.acm.org/citation.cfm?id=3045003","open_access":"1"}],"page":"991 - 999","volume":32,"citation":{"mla":"Pentina, Anastasia, and Christoph Lampert. <i>A PAC-Bayesian Bound for Lifelong Learning</i>. Vol. 32, ML Research Press, 2014, pp. 991–99.","short":"A. Pentina, C. Lampert, in:, ML Research Press, 2014, pp. 991–999.","apa":"Pentina, A., &#38; Lampert, C. (2014). A PAC-Bayesian bound for Lifelong Learning (Vol. 32, pp. 991–999). Presented at the ICML: International Conference on Machine Learning, Beijing, China: ML Research Press.","ieee":"A. Pentina and C. Lampert, “A PAC-Bayesian bound for Lifelong Learning,” presented at the ICML: International Conference on Machine Learning, Beijing, China, 2014, vol. 32, pp. 991–999.","ista":"Pentina A, Lampert C. 2014. A PAC-Bayesian bound for Lifelong Learning. ICML: International Conference on Machine Learning vol. 32, 991–999.","ama":"Pentina A, Lampert C. A PAC-Bayesian bound for Lifelong Learning. In: Vol 32. ML Research Press; 2014:991-999.","chicago":"Pentina, Anastasia, and Christoph Lampert. “A PAC-Bayesian Bound for Lifelong Learning,” 32:991–99. ML Research Press, 2014."},"date_created":"2018-12-11T11:56:03Z","oa_version":"Submitted Version","date_updated":"2024-10-09T20:55:36Z","quality_controlled":"1","scopus_import":"1","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","publisher":"ML Research Press","day":"10","title":"A PAC-Bayesian bound for Lifelong Learning"},{"scopus_import":"1","issue":"PART 3","day":"01","title":"Closed-form approximate CRF training for scalable image segmentation","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","type":"conference","date_created":"2018-12-11T11:56:07Z","citation":{"apa":"Kolesnikov, A., Guillaumin, M., Ferrari, V., &#38; Lampert, C. (2014). Closed-form approximate CRF training for scalable image segmentation. In D. Fleet, T. Pajdla, B. Schiele, &#38; T. Tuytelaars (Eds.), <i>Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)</i> (Vol. 8691, pp. 550–565). Zurich, Switzerland: Springer. <a href=\"https://doi.org/10.1007/978-3-319-10578-9_36\">https://doi.org/10.1007/978-3-319-10578-9_36</a>","ieee":"A. Kolesnikov, M. Guillaumin, V. Ferrari, and C. Lampert, “Closed-form approximate CRF training for scalable image segmentation,” in <i>Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)</i>, Zurich, Switzerland, 2014, vol. 8691, no. PART 3, pp. 550–565.","short":"A. Kolesnikov, M. Guillaumin, V. Ferrari, C. Lampert, in:, D. Fleet, T. Pajdla, B. Schiele, T. Tuytelaars (Eds.), Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Springer, 2014, pp. 550–565.","mla":"Kolesnikov, Alexander, et al. “Closed-Form Approximate CRF Training for Scalable Image Segmentation.” <i>Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)</i>, edited by David Fleet et al., vol. 8691, no. PART 3, Springer, 2014, pp. 550–65, doi:<a href=\"https://doi.org/10.1007/978-3-319-10578-9_36\">10.1007/978-3-319-10578-9_36</a>.","chicago":"Kolesnikov, Alexander, Matthieu Guillaumin, Vittorio Ferrari, and Christoph Lampert. “Closed-Form Approximate CRF Training for Scalable Image Segmentation.” In <i>Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)</i>, edited by David Fleet, Tomas Pajdla, Bernt Schiele, and Tinne Tuytelaars, 8691:550–65. Springer, 2014. <a href=\"https://doi.org/10.1007/978-3-319-10578-9_36\">https://doi.org/10.1007/978-3-319-10578-9_36</a>.","ama":"Kolesnikov A, Guillaumin M, Ferrari V, Lampert C. Closed-form approximate CRF training for scalable image segmentation. In: Fleet D, Pajdla T, Schiele B, Tuytelaars T, eds. <i>Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)</i>. Vol 8691. Springer; 2014:550-565. doi:<a href=\"https://doi.org/10.1007/978-3-319-10578-9_36\">10.1007/978-3-319-10578-9_36</a>","ista":"Kolesnikov A, Guillaumin M, Ferrari V, Lampert C. 2014. Closed-form approximate CRF training for scalable image segmentation. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). ECCV: European Conference on Computer Vision, LNCS, vol. 8691, 550–565."},"abstract":[{"lang":"eng","text":"We present LS-CRF, a new method for training cyclic Conditional Random Fields (CRFs) from large datasets that is inspired by classical closed-form expressions for the maximum likelihood parameters of a generative graphical model with tree topology. Training a CRF with LS-CRF requires only solving a set of independent regression problems, each of which can be solved efficiently in closed form or by an iterative solver. This makes LS-CRF orders of magnitude faster than classical CRF training based on probabilistic inference, and at the same time more flexible and easier to implement than other approximate techniques, such as pseudolikelihood or piecewise training. We apply LS-CRF to the task of semantic image segmentation, showing that it achieves on par accuracy to other training techniques at higher speed, thereby allowing efficient CRF training from very large training sets. For example, training a linearly parameterized pairwise CRF on 150,000 images requires less than one hour on a modern workstation."}],"_id":"2171","status":"public","department":[{"_id":"ChLa"}],"date_published":"2014-09-01T00:00:00Z","oa_version":"Submitted Version","date_updated":"2025-06-11T07:59:20Z","quality_controlled":"1","publisher":"Springer","publication":"Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)","editor":[{"last_name":"Fleet","first_name":"David","full_name":"Fleet, David"},{"full_name":"Pajdla, Tomas","first_name":"Tomas","last_name":"Pajdla"},{"last_name":"Schiele","first_name":"Bernt","full_name":"Schiele, Bernt"},{"full_name":"Tuytelaars, Tinne","last_name":"Tuytelaars","first_name":"Tinne"}],"year":"2014","main_file_link":[{"open_access":"1","url":"http://arxiv.org/abs/1403.7057"}],"external_id":{"arxiv":["1403.7057"]},"page":"550 - 565","volume":8691,"intvolume":"      8691","publication_status":"published","month":"09","conference":{"start_date":"2014-09-06","name":"ECCV: European Conference on Computer Vision","end_date":"2014-09-12","location":"Zurich, Switzerland"},"publist_id":"4813","ec_funded":1,"language":[{"iso":"eng"}],"project":[{"name":"Lifelong Learning of Visual Scene Understanding","grant_number":"308036","_id":"2532554C-B435-11E9-9278-68D0E5697425","call_identifier":"FP7"}],"author":[{"last_name":"Kolesnikov","first_name":"Alexander","id":"2D157DB6-F248-11E8-B48F-1D18A9856A87","full_name":"Kolesnikov, Alexander"},{"full_name":"Guillaumin, Matthieu","last_name":"Guillaumin","first_name":"Matthieu"},{"full_name":"Ferrari, Vittorio","last_name":"Ferrari","first_name":"Vittorio"},{"orcid":"0000-0001-8622-7887","last_name":"Lampert","first_name":"Christoph","id":"40C20FD2-F248-11E8-B48F-1D18A9856A87","full_name":"Lampert, Christoph"}],"alternative_title":["LNCS"],"oa":1,"article_processing_charge":"No","arxiv":1,"fulldoi":"https://doi.org/10.1007/978-3-319-10578-9_36","doi":"10.1007/978-3-319-10578-9_36"},{"abstract":[{"text":"Fisher Kernels and Deep Learning were two developments with significant impact on large-scale object categorization in the last years. Both approaches were shown to achieve state-of-the-art results on large-scale object categorization datasets, such as ImageNet. Conceptually, however, they are perceived as very different and it is not uncommon for heated debates to spring up when advocates of both paradigms meet at conferences or workshops. In this work, we emphasize the similarities between both architectures rather than their differences and we argue that such a unified view allows us to transfer ideas from one domain to the other. As a concrete example we introduce a method for learning a support vector machine classifier with Fisher kernel at the same time as a task-specific data representation. We reinterpret the setting as a multi-layer feed forward network. Its final layer is the classifier, parameterized by a weight vector, and the two previous layers compute Fisher vectors, parameterized by the coefficients of a Gaussian mixture model. We introduce a gradient descent based learning algorithm that, in contrast to other feature learning techniques, is not just derived from intuition or biological analogy, but has a theoretical justification in the framework of statistical learning theory. Our experiments show that the new training procedure leads to significant improvements in classification accuracy while preserving the modularity and geometric interpretability of a support vector machine setup.","lang":"eng"}],"_id":"2172","department":[{"_id":"ChLa"}],"date_published":"2014-09-24T00:00:00Z","status":"public","title":"Deep Fisher Kernels – End to end learning of the Fisher Kernel GMM parameters","day":"24","user_id":"317138e5-6ab7-11ef-aa6d-ffef3953e345","scopus_import":"1","isi":1,"citation":{"ista":"Sydorov V, Sakurada M, Lampert C. 2014. Deep Fisher Kernels – End to end learning of the Fisher Kernel GMM parameters. Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition. CVPR: Computer Vision and Pattern Recognition, 1402–1409.","ama":"Sydorov V, Sakurada M, Lampert C. Deep Fisher Kernels – End to end learning of the Fisher Kernel GMM parameters. In: <i>Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition</i>. IEEE; 2014:1402-1409. doi:<a href=\"https://doi.org/10.1109/CVPR.2014.182\">10.1109/CVPR.2014.182</a>","chicago":"Sydorov, Vladyslav, Mayu Sakurada, and Christoph Lampert. “Deep Fisher Kernels – End to End Learning of the Fisher Kernel GMM Parameters.” In <i>Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition</i>, 1402–9. IEEE, 2014. <a href=\"https://doi.org/10.1109/CVPR.2014.182\">https://doi.org/10.1109/CVPR.2014.182</a>.","mla":"Sydorov, Vladyslav, et al. “Deep Fisher Kernels – End to End Learning of the Fisher Kernel GMM Parameters.” <i>Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition</i>, IEEE, 2014, pp. 1402–09, doi:<a href=\"https://doi.org/10.1109/CVPR.2014.182\">10.1109/CVPR.2014.182</a>.","apa":"Sydorov, V., Sakurada, M., &#38; Lampert, C. (2014). Deep Fisher Kernels – End to end learning of the Fisher Kernel GMM parameters. In <i>Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition</i> (pp. 1402–1409). Columbus, USA: IEEE. <a href=\"https://doi.org/10.1109/CVPR.2014.182\">https://doi.org/10.1109/CVPR.2014.182</a>","short":"V. Sydorov, M. Sakurada, C. Lampert, in:, Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, IEEE, 2014, pp. 1402–1409.","ieee":"V. Sydorov, M. Sakurada, and C. Lampert, “Deep Fisher Kernels – End to end learning of the Fisher Kernel GMM parameters,” in <i>Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition</i>, Columbus, USA, 2014, pp. 1402–1409."},"date_created":"2018-12-11T11:56:08Z","type":"conference","project":[{"call_identifier":"FP7","_id":"2532554C-B435-11E9-9278-68D0E5697425","name":"Lifelong Learning of Visual Scene Understanding","grant_number":"308036"}],"ec_funded":1,"language":[{"iso":"eng"}],"author":[{"last_name":"Sydorov","first_name":"Vladyslav","full_name":"Sydorov, Vladyslav"},{"full_name":"Sakurada, Mayu","last_name":"Sakurada","first_name":"Mayu"},{"full_name":"Lampert, Christoph","last_name":"Lampert","first_name":"Christoph","orcid":"0000-0001-8622-7887","id":"40C20FD2-F248-11E8-B48F-1D18A9856A87"}],"publication_status":"published","month":"09","publist_id":"4812","conference":{"location":"Columbus, USA","end_date":"2014-06-28","name":"CVPR: Computer Vision and Pattern Recognition","start_date":"2014-06-23"},"doi":"10.1109/CVPR.2014.182","corr_author":"1","fulldoi":"https://doi.org/10.1109/CVPR.2014.182","article_processing_charge":"No","publisher":"IEEE","publication":"Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition","quality_controlled":"1","date_updated":"2025-09-29T11:38:10Z","oa_version":"None","page":"1402 - 1409","external_id":{"isi":["000361555601057"]},"year":"2014"},{"year":"2014","file":[{"date_updated":"2020-07-14T12:45:31Z","checksum":"c4c6d3efdb8ee648faf3e76849839ce2","file_id":"4683","file_size":408172,"file_name":"IST-2016-490-v1+1_khamis-bmvc2014.pdf","access_level":"open_access","content_type":"application/pdf","creator":"system","relation":"main_file","date_created":"2018-12-12T10:08:23Z"}],"date_updated":"2021-01-12T06:55:46Z","quality_controlled":"1","oa_version":"Published Version","publication":"Proceedings of the British Machine Vision Conference 2014","publisher":"BMVA Press","oa":1,"pubrep_id":"490","publist_id":"4811","conference":{"end_date":"2014-09-05","start_date":"2014-09-01","name":"BMVC: British Machine Vision Conference","location":"Nottingham, UK"},"ddc":["000"],"publication_status":"published","month":"09","author":[{"first_name":"Sameh","last_name":"Khamis","full_name":"Khamis, Sameh"},{"orcid":"0000-0001-8622-7887","first_name":"Christoph","last_name":"Lampert","id":"40C20FD2-F248-11E8-B48F-1D18A9856A87","full_name":"Lampert, Christoph"}],"project":[{"name":"Lifelong Learning of Visual Scene Understanding","grant_number":"308036","call_identifier":"FP7","_id":"2532554C-B435-11E9-9278-68D0E5697425"}],"language":[{"iso":"eng"}],"ec_funded":1,"type":"conference","citation":{"chicago":"Khamis, Sameh, and Christoph Lampert. “CoConut: Co-Classification with Output Space Regularization.” In <i>Proceedings of the British Machine Vision Conference 2014</i>. BMVA Press, 2014.","ama":"Khamis S, Lampert C. CoConut: Co-classification with output space regularization. In: <i>Proceedings of the British Machine Vision Conference 2014</i>. BMVA Press; 2014.","ista":"Khamis S, Lampert C. 2014. CoConut: Co-classification with output space regularization. Proceedings of the British Machine Vision Conference 2014. BMVC: British Machine Vision Conference.","short":"S. Khamis, C. Lampert, in:, Proceedings of the British Machine Vision Conference 2014, BMVA Press, 2014.","apa":"Khamis, S., &#38; Lampert, C. (2014). CoConut: Co-classification with output space regularization. In <i>Proceedings of the British Machine Vision Conference 2014</i>. Nottingham, UK: BMVA Press.","ieee":"S. Khamis and C. Lampert, “CoConut: Co-classification with output space regularization,” in <i>Proceedings of the British Machine Vision Conference 2014</i>, Nottingham, UK, 2014.","mla":"Khamis, Sameh, and Christoph Lampert. “CoConut: Co-Classification with Output Space Regularization.” <i>Proceedings of the British Machine Vision Conference 2014</i>, BMVA Press, 2014."},"date_created":"2018-12-11T11:56:08Z","scopus_import":1,"user_id":"3E5EF7F0-F248-11E8-B48F-1D18A9856A87","title":"CoConut: Co-classification with output space regularization","day":"01","status":"public","date_published":"2014-09-01T00:00:00Z","department":[{"_id":"ChLa"}],"file_date_updated":"2020-07-14T12:45:31Z","has_accepted_license":"1","_id":"2173","abstract":[{"lang":"eng","text":"In this work we introduce a new approach to co-classification, i.e. the task of jointly classifying multiple, otherwise independent, data samples. The method we present, named CoConut, is based on the idea of adding a regularizer in the label space to encode certain priors on the resulting labelings. A regularizer that encourages labelings that are smooth across the test set, for instance, can be seen as a test-time variant of the cluster assumption, which has been proven useful at training time in semi-supervised learning. A regularizer that introduces a preference for certain class proportions can be regarded as a prior distribution on the class labels. CoConut can build on existing classifiers without making any assumptions on how they were obtained and without the need to re-train them. The use of a regularizer adds a new level of flexibility. It allows the integration of potentially new information at test time, even in other modalities than what the classifiers were trained on. We evaluate our framework on six datasets, reporting a clear performance gain in classification accuracy compared to the standard classification setup that predicts labels for each test sample separately.\r\n"}]},{"main_file_link":[{"url":"https://hal.archives-ouvertes.fr/hal-01009578/document","open_access":"1"}],"external_id":{"isi":["000341431300007"]},"page":"129 - 154","volume":97,"year":"2014","publication":"Machine Learning","publisher":"Springer","quality_controlled":"1","date_updated":"2025-09-29T11:35:24Z","oa_version":"Submitted Version","doi":"10.1007/s10994-014-5462-z","oa":1,"corr_author":"1","fulldoi":"https://doi.org/10.1007/s10994-014-5462-z","article_processing_charge":"No","project":[{"grant_number":"308036","name":"Lifelong Learning of Visual Scene Understanding","call_identifier":"FP7","_id":"2532554C-B435-11E9-9278-68D0E5697425"}],"ec_funded":1,"language":[{"iso":"eng"}],"author":[{"last_name":"Bellet","first_name":"Aurélien","full_name":"Bellet, Aurélien"},{"last_name":"Habrard","first_name":"Amaury","full_name":"Habrard, Amaury"},{"last_name":"Morvant","first_name":"Emilie","orcid":"0000-0002-8301-7240","id":"4BAC2A72-F248-11E8-B48F-1D18A9856A87","full_name":"Morvant, Emilie"},{"last_name":"Sebban","first_name":"Marc","full_name":"Sebban, Marc"}],"month":"10","intvolume":"        97","publication_status":"published","publist_id":"4802","date_created":"2018-12-11T11:56:10Z","citation":{"short":"A. Bellet, A. Habrard, E. Morvant, M. Sebban, Machine Learning 97 (2014) 129–154.","apa":"Bellet, A., Habrard, A., Morvant, E., &#38; Sebban, M. (2014). Learning a priori constrained weighted majority votes. <i>Machine Learning</i>. Springer. <a href=\"https://doi.org/10.1007/s10994-014-5462-z\">https://doi.org/10.1007/s10994-014-5462-z</a>","ieee":"A. Bellet, A. Habrard, E. Morvant, and M. Sebban, “Learning a priori constrained weighted majority votes,” <i>Machine Learning</i>, vol. 97, no. 1–2. Springer, pp. 129–154, 2014.","mla":"Bellet, Aurélien, et al. “Learning a Priori Constrained Weighted Majority Votes.” <i>Machine Learning</i>, vol. 97, no. 1–2, Springer, 2014, pp. 129–54, doi:<a href=\"https://doi.org/10.1007/s10994-014-5462-z\">10.1007/s10994-014-5462-z</a>.","ista":"Bellet A, Habrard A, Morvant E, Sebban M. 2014. Learning a priori constrained weighted majority votes. Machine Learning. 97(1–2), 129–154.","ama":"Bellet A, Habrard A, Morvant E, Sebban M. Learning a priori constrained weighted majority votes. <i>Machine Learning</i>. 2014;97(1-2):129-154. doi:<a href=\"https://doi.org/10.1007/s10994-014-5462-z\">10.1007/s10994-014-5462-z</a>","chicago":"Bellet, Aurélien, Amaury Habrard, Emilie Morvant, and Marc Sebban. “Learning a Priori Constrained Weighted Majority Votes.” <i>Machine Learning</i>. Springer, 2014. <a href=\"https://doi.org/10.1007/s10994-014-5462-z\">https://doi.org/10.1007/s10994-014-5462-z</a>."},"type":"journal_article","title":"Learning a priori constrained weighted majority votes","day":"01","user_id":"317138e5-6ab7-11ef-aa6d-ffef3953e345","scopus_import":"1","isi":1,"issue":"1-2","department":[{"_id":"ChLa"}],"date_published":"2014-10-01T00:00:00Z","status":"public","_id":"2180","abstract":[{"text":"Weighted majority votes allow one to combine the output of several classifiers or voters. MinCq is a recent algorithm for optimizing the weight of each voter based on the minimization of a theoretical bound over the risk of the vote with elegant PAC-Bayesian generalization guarantees. However, while it has demonstrated good performance when combining weak classifiers, MinCq cannot make use of the useful a priori knowledge that one may have when using a mixture of weak and strong voters. In this paper, we propose P-MinCq, an extension of MinCq that can incorporate such knowledge in the form of a  constraint over the distribution of the weights, along with general proofs of convergence that stand in the sample compression setting for data-dependent voters. The approach is applied to a vote of k-NN classifiers with a specific modeling of the voters' performance. P-MinCq significantly outperforms the classic k-NN classifier, a symmetric NN and MinCq using the same voters. We show that it is also competitive with LMNN, a popular metric learning algorithm, and that combining both approaches further reduces the error.","lang":"eng"}],"acknowledgement":"This work was funded by the French project SoLSTiCe ANR-13-BS02-01 of the ANR. "},{"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","day":"01","publisher":"Elsevier","title":"Adaptation de domaine de vote de majorité par auto-étiquetage non itératif","oa_version":"Preprint","quality_controlled":"1","date_updated":"2024-10-09T20:55:32Z","page":"49-58","main_file_link":[{"url":"https://hal.archives-ouvertes.fr/hal-01005776/","open_access":"1"}],"volume":1,"citation":{"ama":"Morvant E. Adaptation de domaine de vote de majorité par auto-étiquetage non itératif. In: Vol 1. Elsevier; 2014:49-58.","chicago":"Morvant, Emilie. “Adaptation de Domaine de Vote de Majorité Par Auto-Étiquetage Non Itératif,” 1:49–58. Elsevier, 2014.","ista":"Morvant E. 2014. Adaptation de domaine de vote de majorité par auto-étiquetage non itératif. CAP: Conférence Francophone sur l’Apprentissage Automatique (Machine Learning French Conference) vol. 1, 49–58.","mla":"Morvant, Emilie. <i>Adaptation de Domaine de Vote de Majorité Par Auto-Étiquetage Non Itératif</i>. Vol. 1, Elsevier, 2014, pp. 49–58.","short":"E. Morvant, in:, Elsevier, 2014, pp. 49–58.","apa":"Morvant, E. (2014). Adaptation de domaine de vote de majorité par auto-étiquetage non itératif (Vol. 1, pp. 49–58). Presented at the CAP: Conférence Francophone sur l’Apprentissage Automatique (Machine Learning French Conference), Saint-Etienne, France: Elsevier.","ieee":"E. Morvant, “Adaptation de domaine de vote de majorité par auto-étiquetage non itératif,” presented at the CAP: Conférence Francophone sur l’Apprentissage Automatique (Machine Learning French Conference), Saint-Etienne, France, 2014, vol. 1, pp. 49–58."},"date_created":"2018-12-11T11:56:13Z","type":"conference","year":"2014","abstract":[{"lang":"fre","text":"En apprentissage automatique, nous parlons d'adaptation de domaine lorsque les données de test (cibles) et d'apprentissage (sources) sont générées selon différentes distributions. Nous devons donc développer des algorithmes de classification capables de s'adapter à une nouvelle distribution, pour laquelle aucune information sur les étiquettes n'est disponible. Nous attaquons cette problématique sous l'angle de l'approche PAC-Bayésienne qui se focalise sur l'apprentissage de modèles définis comme des votes de majorité sur un ensemble de fonctions. Dans ce contexte, nous introduisons PV-MinCq une version adaptative de l'algorithme (non adaptatif) MinCq. PV-MinCq suit le principe suivant. Nous transférons les étiquettes sources aux points cibles proches pour ensuite appliquer MinCq sur l'échantillon cible ``auto-étiqueté'' (justifié par une borne théorique). Plus précisément, nous définissons un auto-étiquetage non itératif qui se focalise dans les régions où les distributions marginales source et cible sont les plus similaires. Dans un second temps, nous étudions l'influence de notre auto-étiquetage pour en déduire une procédure de validation des hyperparamètres. Finalement, notre approche montre des résultats empiriques prometteurs."}],"_id":"2189","author":[{"full_name":"Morvant, Emilie","id":"4BAC2A72-F248-11E8-B48F-1D18A9856A87","first_name":"Emilie","last_name":"Morvant","orcid":"0000-0002-8301-7240"}],"language":[{"iso":"eng"}],"conference":{"name":"CAP: Conférence Francophone sur l'Apprentissage Automatique (Machine Learning French Conference)","location":"Saint-Etienne, France"},"publist_id":"4785","publication_status":"published","intvolume":"         1","month":"07","date_published":"2014-07-01T00:00:00Z","department":[{"_id":"ChLa"}],"article_processing_charge":"No","corr_author":"1","oa":1,"status":"public"},{"volume":1,"page":"837-845","main_file_link":[{"open_access":"1","url":"https://papers.nips.cc/paper/5373-mind-the-nuisance-gaussian-process-classification-using-privileged-noise"}],"year":"2014","publication":"Advances in Neural Information Processing Systems","publisher":"Neural Information Processing Systems Foundation","date_updated":"2025-06-03T11:45:29Z","quality_controlled":"1","oa_version":"Submitted Version","article_processing_charge":"No","oa":1,"author":[{"first_name":"Daniel","last_name":"Hernandez Lobato","full_name":"Hernandez Lobato, Daniel"},{"full_name":"Sharmanska, Viktoriia","id":"2EA6D09E-F248-11E8-B48F-1D18A9856A87","last_name":"Sharmanska","first_name":"Viktoriia","orcid":"0000-0003-0192-9308"},{"last_name":"Kersting","first_name":"Kristian","full_name":"Kersting, Kristian"},{"last_name":"Lampert","first_name":"Christoph","orcid":"0000-0001-8622-7887","id":"40C20FD2-F248-11E8-B48F-1D18A9856A87","full_name":"Lampert, Christoph"},{"first_name":"Novi","last_name":"Quadrianto","full_name":"Quadrianto, Novi"}],"language":[{"iso":"eng"}],"publist_id":"5038","conference":{"end_date":"2014-12-13","start_date":"2014-12-08","name":"NIPS: Neural Information Processing Systems","location":"Montreal, Canada"},"month":"12","intvolume":"         1","publication_status":"published","citation":{"chicago":"Hernandez Lobato, Daniel, Viktoriia Sharmanska, Kristian Kersting, Christoph Lampert, and Novi Quadrianto. “Mind the Nuisance: Gaussian Process Classification Using Privileged Noise.” In <i>Advances in Neural Information Processing Systems</i>, 1:837–45. Neural Information Processing Systems Foundation, 2014.","ama":"Hernandez Lobato D, Sharmanska V, Kersting K, Lampert C, Quadrianto N. Mind the nuisance: Gaussian process classification using privileged noise. In: <i>Advances in Neural Information Processing Systems</i>. Vol 1. Neural Information Processing Systems Foundation; 2014:837-845.","ista":"Hernandez Lobato D, Sharmanska V, Kersting K, Lampert C, Quadrianto N. 2014. Mind the nuisance: Gaussian process classification using privileged noise. Advances in Neural Information Processing Systems. NIPS: Neural Information Processing Systems vol. 1, 837–845.","ieee":"D. Hernandez Lobato, V. Sharmanska, K. Kersting, C. Lampert, and N. Quadrianto, “Mind the nuisance: Gaussian process classification using privileged noise,” in <i>Advances in Neural Information Processing Systems</i>, Montreal, Canada, 2014, vol. 1, no. January, pp. 837–845.","short":"D. Hernandez Lobato, V. Sharmanska, K. Kersting, C. Lampert, N. Quadrianto, in:, Advances in Neural Information Processing Systems, Neural Information Processing Systems Foundation, 2014, pp. 837–845.","apa":"Hernandez Lobato, D., Sharmanska, V., Kersting, K., Lampert, C., &#38; Quadrianto, N. (2014). Mind the nuisance: Gaussian process classification using privileged noise. In <i>Advances in Neural Information Processing Systems</i> (Vol. 1, pp. 837–845). Montreal, Canada: Neural Information Processing Systems Foundation.","mla":"Hernandez Lobato, Daniel, et al. “Mind the Nuisance: Gaussian Process Classification Using Privileged Noise.” <i>Advances in Neural Information Processing Systems</i>, vol. 1, no. January, Neural Information Processing Systems Foundation, 2014, pp. 837–45."},"date_created":"2018-12-11T11:55:20Z","type":"conference","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","title":"Mind the nuisance: Gaussian process classification using privileged noise","day":"08","issue":"January","scopus_import":"1","date_published":"2014-12-08T00:00:00Z","department":[{"_id":"ChLa"}],"status":"public","abstract":[{"text":"The learning with privileged information setting has recently attracted a lot of attention within the machine learning community, as it allows the integration of additional knowledge into the training process of a classifier, even when this comes in the form of a data modality that is not available at test time. Here, we show that privileged information can naturally be treated as noise in the latent function of a Gaussian process classifier (GPC). That is, in contrast to the standard GPC setting, the latent function is not just a nuisance but a feature: it becomes a natural measure of confidence about the training data by modulating the slope of the GPC probit likelihood function. Extensive experiments on public datasets show that the proposed GPC method using privileged noise, called GPC+, improves over a standard GPC without privileged knowledge, and also over the current state-of-the-art SVM-based method, SVM+. Moreover, we show that advanced neural networks and deep learning methods can be compressed as privileged information.","lang":"eng"}],"_id":"2033"},{"publication_status":"published","month":"01","intvolume":"        27","ddc":["000"],"OA_type":"gold","conference":{"end_date":"2014-12-13","name":"NIPS: Neural Information Processing Systems","start_date":"2014-12-08","location":"Montreal, Canada"},"publist_id":"4996","language":[{"iso":"eng"}],"author":[{"full_name":"Marchand, Mario","last_name":"Marchand","first_name":"Mario"},{"full_name":"Hongyu, Su","last_name":"Hongyu","first_name":"Su"},{"full_name":"Morvant, Emilie","id":"4BAC2A72-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0002-8301-7240","first_name":"Emilie","last_name":"Morvant"},{"full_name":"Rousu, Juho","last_name":"Rousu","first_name":"Juho"},{"first_name":"John","last_name":"Shawe Taylor","full_name":"Shawe Taylor, John"}],"oa":1,"publication_identifier":{"isbn":["9781510800410"]},"article_processing_charge":"No","oa_version":"Published Version","date_updated":"2026-06-18T18:24:19Z","OA_place":"repository","publisher":"Neural Information Processing Systems Foundation","publication":"Advances in Neural Information Processing Systems","year":"2014","volume":27,"main_file_link":[{"url":"https://hal.archives-ouvertes.fr/hal-01065586","open_access":"1"}],"abstract":[{"text":"We show that the usual score function for conditional Markov networks can be written as the expectation over the scores of their spanning trees. We also show that a small random sample of these output trees can attain a significant fraction of the margin obtained by the complete graph and we provide conditions under which we can perform tractable inference. The experimental results confirm that practical learning is scalable to realistic datasets using this approach.","lang":"eng"}],"_id":"2051","status":"public","department":[{"_id":"ChLa"}],"date_published":"2014-01-01T00:00:00Z","day":"01","title":"Multilabel structured output learning with random spanning trees of max-margin Markov networks","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","type":"conference","citation":{"chicago":"Marchand, Mario, Su Hongyu, Emilie Morvant, Juho Rousu, and John Shawe Taylor. “Multilabel Structured Output Learning with Random Spanning Trees of Max-Margin Markov Networks.” In <i>Advances in Neural Information Processing Systems</i>, Vol. 27. Neural Information Processing Systems Foundation, 2014.","ama":"Marchand M, Hongyu S, Morvant E, Rousu J, Shawe Taylor J. Multilabel structured output learning with random spanning trees of max-margin Markov networks. In: <i>Advances in Neural Information Processing Systems</i>. Vol 27. Neural Information Processing Systems Foundation; 2014.","ista":"Marchand M, Hongyu S, Morvant E, Rousu J, Shawe Taylor J. 2014. Multilabel structured output learning with random spanning trees of max-margin Markov networks. Advances in Neural Information Processing Systems. NIPS: Neural Information Processing Systems vol. 27.","ieee":"M. Marchand, S. Hongyu, E. Morvant, J. Rousu, and J. Shawe Taylor, “Multilabel structured output learning with random spanning trees of max-margin Markov networks,” in <i>Advances in Neural Information Processing Systems</i>, Montreal, Canada, 2014, vol. 27.","short":"M. Marchand, S. Hongyu, E. Morvant, J. Rousu, J. Shawe Taylor, in:, Advances in Neural Information Processing Systems, Neural Information Processing Systems Foundation, 2014.","apa":"Marchand, M., Hongyu, S., Morvant, E., Rousu, J., &#38; Shawe Taylor, J. (2014). Multilabel structured output learning with random spanning trees of max-margin Markov networks. In <i>Advances in Neural Information Processing Systems</i> (Vol. 27). Montreal, Canada: Neural Information Processing Systems Foundation.","mla":"Marchand, Mario, et al. “Multilabel Structured Output Learning with Random Spanning Trees of Max-Margin Markov Networks.” <i>Advances in Neural Information Processing Systems</i>, vol. 27, Neural Information Processing Systems Foundation, 2014."},"date_created":"2018-12-11T11:55:26Z"},{"publist_id":"4989","conference":{"location":"Joensuu, Finland","start_date":"2014-08-20","name":"IAPR: International Workshop on Structural, Syntactic, and Statistical Pattern Recognition","end_date":"2014-08-22"},"month":"01","publication_status":"published","intvolume":"      8621","author":[{"full_name":"Morvant, Emilie","id":"4BAC2A72-F248-11E8-B48F-1D18A9856A87","last_name":"Morvant","first_name":"Emilie","orcid":"0000-0002-8301-7240"},{"first_name":"Amaury","last_name":"Habrard","full_name":"Habrard, Amaury"},{"full_name":"Ayache, Stéphane","last_name":"Ayache","first_name":"Stéphane"}],"project":[{"call_identifier":"FP7","_id":"2532554C-B435-11E9-9278-68D0E5697425","grant_number":"308036","name":"Lifelong Learning of Visual Scene Understanding"}],"ec_funded":1,"language":[{"iso":"eng"}],"fulldoi":"https://doi.org/10.1007/978-3-662-44415-3_16","arxiv":1,"oa":1,"alternative_title":["LNCS"],"doi":"10.1007/978-3-662-44415-3_16","date_updated":"2021-01-12T06:55:01Z","quality_controlled":"1","oa_version":"Preprint","publisher":"Springer","publication":"Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)","year":"2014","page":"153 - 162","external_id":{"arxiv":["1404.7796"]},"volume":8621,"main_file_link":[{"open_access":"1","url":"http://arxiv.org/abs/1404.7796"}],"abstract":[{"lang":"eng","text":"In the past few years, a lot of attention has been devoted to multimedia indexing by fusing multimodal informations. Two kinds of fusion schemes are generally considered: The early fusion and the late fusion. We focus on late classifier fusion, where one combines the scores of each modality at the decision level. To tackle this problem, we investigate a recent and elegant well-founded quadratic program named MinCq coming from the machine learning PAC-Bayesian theory. MinCq looks for the weighted combination, over a set of real-valued functions seen as voters, leading to the lowest misclassification rate, while maximizing the voters’ diversity. We propose an extension of MinCq tailored to multimedia indexing. Our method is based on an order-preserving pairwise loss adapted to ranking that allows us to improve Mean Averaged Precision measure while taking into account the diversity of the voters that we want to fuse. We provide evidence that this method is naturally adapted to late fusion procedures and confirm the good behavior of our approach on the challenging PASCAL VOC’07 benchmark."}],"_id":"2057","status":"public","date_published":"2014-01-01T00:00:00Z","department":[{"_id":"ChLa"}],"scopus_import":1,"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","title":"Majority vote of diverse classifiers for late fusion","day":"01","type":"conference","citation":{"ama":"Morvant E, Habrard A, Ayache S. Majority vote of diverse classifiers for late fusion. In: <i>Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)</i>. Vol 8621. Springer; 2014:153-162. doi:<a href=\"https://doi.org/10.1007/978-3-662-44415-3_16\">10.1007/978-3-662-44415-3_16</a>","chicago":"Morvant, Emilie, Amaury Habrard, and Stéphane Ayache. “Majority Vote of Diverse Classifiers for Late Fusion.” In <i>Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)</i>, 8621:153–62. Springer, 2014. <a href=\"https://doi.org/10.1007/978-3-662-44415-3_16\">https://doi.org/10.1007/978-3-662-44415-3_16</a>.","ista":"Morvant E, Habrard A, Ayache S. 2014. Majority vote of diverse classifiers for late fusion. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). IAPR: International Workshop on Structural, Syntactic, and Statistical Pattern Recognition, LNCS, vol. 8621, 153–162.","apa":"Morvant, E., Habrard, A., &#38; Ayache, S. (2014). Majority vote of diverse classifiers for late fusion. In <i>Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)</i> (Vol. 8621, pp. 153–162). Joensuu, Finland: Springer. <a href=\"https://doi.org/10.1007/978-3-662-44415-3_16\">https://doi.org/10.1007/978-3-662-44415-3_16</a>","short":"E. Morvant, A. Habrard, S. Ayache, in:, Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Springer, 2014, pp. 153–162.","ieee":"E. Morvant, A. Habrard, and S. Ayache, “Majority vote of diverse classifiers for late fusion,” in <i>Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)</i>, Joensuu, Finland, 2014, vol. 8621, pp. 153–162.","mla":"Morvant, Emilie, et al. “Majority Vote of Diverse Classifiers for Late Fusion.” <i>Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)</i>, vol. 8621, Springer, 2014, pp. 153–62, doi:<a href=\"https://doi.org/10.1007/978-3-662-44415-3_16\">10.1007/978-3-662-44415-3_16</a>."},"date_created":"2018-12-11T11:55:28Z"},{"type":"book_chapter","citation":{"ama":"Muelling K, Kroemer O, Lampert C, Schölkopf B. Movement templates for learning of hitting and batting. In: Kober J, Peters J, eds. <i>Learning Motor Skills</i>. Vol 97. From Algorithms to Robot Experiments. Springer; 2014:69-82. doi:<a href=\"https://doi.org/10.1007/978-3-319-03194-1_3\">10.1007/978-3-319-03194-1_3</a>","chicago":"Muelling, Katharina, Oliver Kroemer, Christoph Lampert, and Bernhard Schölkopf. “Movement Templates for Learning of Hitting and Batting.” In <i>Learning Motor Skills</i>, edited by Jens Kober and Jan Peters, 97:69–82. From Algorithms to Robot Experiments. Springer, 2014. <a href=\"https://doi.org/10.1007/978-3-319-03194-1_3\">https://doi.org/10.1007/978-3-319-03194-1_3</a>.","ista":"Muelling K, Kroemer O, Lampert C, Schölkopf B. 2014.Movement templates for learning of hitting and batting. In: Learning Motor Skills. Springer Tracts in Advanced Robotics, vol. 97, 69–82.","ieee":"K. Muelling, O. Kroemer, C. Lampert, and B. Schölkopf, “Movement templates for learning of hitting and batting,” in <i>Learning Motor Skills</i>, vol. 97, J. Kober and J. Peters, Eds. Springer, 2014, pp. 69–82.","short":"K. Muelling, O. Kroemer, C. Lampert, B. Schölkopf, in:, J. Kober, J. Peters (Eds.), Learning Motor Skills, Springer, 2014, pp. 69–82.","apa":"Muelling, K., Kroemer, O., Lampert, C., &#38; Schölkopf, B. (2014). Movement templates for learning of hitting and batting. In J. Kober &#38; J. Peters (Eds.), <i>Learning Motor Skills</i> (Vol. 97, pp. 69–82). Springer. <a href=\"https://doi.org/10.1007/978-3-319-03194-1_3\">https://doi.org/10.1007/978-3-319-03194-1_3</a>","mla":"Muelling, Katharina, et al. “Movement Templates for Learning of Hitting and Batting.” <i>Learning Motor Skills</i>, edited by Jens Kober and Jan Peters, vol. 97, Springer, 2014, pp. 69–82, doi:<a href=\"https://doi.org/10.1007/978-3-319-03194-1_3\">10.1007/978-3-319-03194-1_3</a>."},"date_created":"2018-12-11T11:54:14Z","series_title":"From Algorithms to Robot Experiments","scopus_import":1,"user_id":"4435EBFC-F248-11E8-B48F-1D18A9856A87","day":"01","title":"Movement templates for learning of hitting and batting","status":"public","date_published":"2014-01-01T00:00:00Z","department":[{"_id":"ChLa"}],"_id":"1829","abstract":[{"lang":"eng","text":"Hitting and batting tasks, such as tennis forehands, ping-pong strokes, or baseball batting, depend on predictions where the ball can be intercepted and how it can properly be returned to the opponent. These predictions get more accurate over time, hence the behaviors need to be continuously modified. As a result, movement templates with a learned global shape need to be adapted during the execution so that the racket reaches a target position and velocity that will return the ball over to the other side of the net or court. It requires altering learned movements to hit a varying target with the necessary velocity at a specific instant in time. Such a task cannot be incorporated straightforwardly in most movement representations suitable for learning. For example, the standard formulation of the dynamical system based motor primitives (introduced by Ijspeert et al (2002b)) does not satisfy this property despite their flexibility which has allowed learning tasks ranging from locomotion to kendama. In order to fulfill this requirement, we reformulate the Ijspeert framework to incorporate the possibility of specifying a desired hitting point and a desired hitting velocity while maintaining all advantages of the original formulation.We show that the proposed movement template formulation works well in two scenarios, i.e., for hitting a ball on a string with a table tennis racket at a specified velocity and for returning balls launched by a ball gun successfully over the net using forehand movements."}],"year":"2014","volume":97,"page":"69 - 82","oa_version":"None","date_updated":"2021-01-12T06:53:28Z","quality_controlled":"1","editor":[{"last_name":"Kober","first_name":"Jens","full_name":"Kober, Jens"},{"last_name":"Peters","first_name":"Jan","full_name":"Peters, Jan"}],"publisher":"Springer","publication":"Learning Motor Skills","fulldoi":"https://doi.org/10.1007/978-3-319-03194-1_3","alternative_title":["Springer Tracts in Advanced Robotics"],"doi":"10.1007/978-3-319-03194-1_3","publist_id":"5274","month":"01","intvolume":"        97","publication_status":"published","author":[{"first_name":"Katharina","last_name":"Muelling","full_name":"Muelling, Katharina"},{"last_name":"Kroemer","first_name":"Oliver","full_name":"Kroemer, Oliver"},{"full_name":"Lampert, Christoph","id":"40C20FD2-F248-11E8-B48F-1D18A9856A87","first_name":"Christoph","last_name":"Lampert","orcid":"0000-0001-8622-7887"},{"full_name":"Schölkopf, Bernhard","last_name":"Schölkopf","first_name":"Bernhard"}],"language":[{"iso":"eng"}]},{"external_id":{"isi":["000351830500103"]},"main_file_link":[{"open_access":"1","url":"www.cv-foundation.org/openaccess/content_iccv_2013/papers/Sharmanska_Learning_to_Rank_2013_ICCV_paper.pdf"}],"page":"825 - 832","year":"2013","publisher":"IEEE","oa_version":"Submitted Version","quality_controlled":"1","date_updated":"2025-09-29T14:20:45Z","doi":"10.1109/ICCV.2013.107","oa":1,"article_processing_charge":"No","fulldoi":"https://doi.org/10.1109/ICCV.2013.107","ec_funded":1,"language":[{"iso":"eng"}],"project":[{"grant_number":"308036","name":"Lifelong Learning of Visual Scene Understanding","call_identifier":"FP7","_id":"2532554C-B435-11E9-9278-68D0E5697425"}],"author":[{"full_name":"Sharmanska, Viktoriia","id":"2EA6D09E-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0003-0192-9308","last_name":"Sharmanska","first_name":"Viktoriia"},{"full_name":"Quadrianto, Novi","last_name":"Quadrianto","first_name":"Novi"},{"full_name":"Lampert, Christoph","id":"40C20FD2-F248-11E8-B48F-1D18A9856A87","first_name":"Christoph","last_name":"Lampert","orcid":"0000-0001-8622-7887"}],"month":"12","publication_status":"published","conference":{"name":"ICCV: International Conference on Computer Vision","start_date":"2013-12-01","end_date":"2013-12-08","location":"Sydney, Australia"},"publist_id":"4635","date_created":"2018-12-11T11:56:49Z","citation":{"chicago":"Sharmanska, Viktoriia, Novi Quadrianto, and Christoph Lampert. “Learning to Rank Using Privileged Information,” 825–32. IEEE, 2013. <a href=\"https://doi.org/10.1109/ICCV.2013.107\">https://doi.org/10.1109/ICCV.2013.107</a>.","ama":"Sharmanska V, Quadrianto N, Lampert C. Learning to rank using privileged information. In: IEEE; 2013:825-832. doi:<a href=\"https://doi.org/10.1109/ICCV.2013.107\">10.1109/ICCV.2013.107</a>","ista":"Sharmanska V, Quadrianto N, Lampert C. 2013. Learning to rank using privileged information. ICCV: International Conference on Computer Vision, 825–832.","mla":"Sharmanska, Viktoriia, et al. <i>Learning to Rank Using Privileged Information</i>. IEEE, 2013, pp. 825–32, doi:<a href=\"https://doi.org/10.1109/ICCV.2013.107\">10.1109/ICCV.2013.107</a>.","apa":"Sharmanska, V., Quadrianto, N., &#38; Lampert, C. (2013). Learning to rank using privileged information (pp. 825–832). Presented at the ICCV: International Conference on Computer Vision, Sydney, Australia: IEEE. <a href=\"https://doi.org/10.1109/ICCV.2013.107\">https://doi.org/10.1109/ICCV.2013.107</a>","short":"V. Sharmanska, N. Quadrianto, C. Lampert, in:, IEEE, 2013, pp. 825–832.","ieee":"V. Sharmanska, N. Quadrianto, and C. Lampert, “Learning to rank using privileged information,” presented at the ICCV: International Conference on Computer Vision, Sydney, Australia, 2013, pp. 825–832."},"type":"conference","day":"01","title":"Learning to rank using privileged information","user_id":"317138e5-6ab7-11ef-aa6d-ffef3953e345","isi":1,"scopus_import":"1","department":[{"_id":"ChLa"}],"date_published":"2013-12-01T00:00:00Z","status":"public","_id":"2293","abstract":[{"text":"Many computer vision problems have an asymmetric distribution of information between training and test time. In this work, we study the case where we are given additional information about the training data, which however will not be available at test time. This situation is called learning using privileged information (LUPI). We introduce two maximum-margin techniques that are able to make use of this additional source of information, and we show that the framework is applicable to several scenarios that have been studied in computer vision before. Experiments with attributes, bounding boxes, image tags and rationales as additional information in object classification show promising results.","lang":"eng"}]},{"doi":"10.1109/ICCV.2013.139","fulldoi":"https://doi.org/10.1109/ICCV.2013.139","article_processing_charge":"No","oa":1,"author":[{"full_name":"Kazmar, Tomas","last_name":"Kazmar","first_name":"Tomas"},{"first_name":"Evgeny","last_name":"Kvon","full_name":"Kvon, Evgeny"},{"last_name":"Stark","first_name":"Alexander","full_name":"Stark, Alexander"},{"last_name":"Lampert","first_name":"Christoph","orcid":"0000-0001-8622-7887","id":"40C20FD2-F248-11E8-B48F-1D18A9856A87","full_name":"Lampert, Christoph"}],"project":[{"name":"Lifelong Learning of Visual Scene Understanding","grant_number":"308036","call_identifier":"FP7","_id":"2532554C-B435-11E9-9278-68D0E5697425"}],"language":[{"iso":"eng"}],"ec_funded":1,"publist_id":"4634","conference":{"location":"Sydney, Australia","end_date":"2013-12-08","start_date":"2013-12-01","name":"ICCV: International Conference on Computer Vision"},"publication_status":"published","month":"12","main_file_link":[{"open_access":"1","url":"http://www.cv-foundation.org/openaccess/ICCV2013.py"}],"external_id":{"isi":["000351830500136"]},"year":"2013","publisher":"IEEE","date_updated":"2025-09-29T14:20:15Z","quality_controlled":"1","oa_version":"Submitted Version","date_published":"2013-12-01T00:00:00Z","department":[{"_id":"ChLa"}],"status":"public","abstract":[{"lang":"eng","text":"In this work we propose a system for automatic classification of Drosophila embryos into developmental stages.\r\nWhile the system is designed to solve an actual problem in biological research, we believe that the principle underly-\r\ning it is interesting not only for biologists, but also for researchers in computer vision. The main idea is to combine two orthogonal sources of information:  one is a classifier trained on strongly invariant features,  which makes it applicable to images of very different conditions, but also leads to rather noisy predictions. The other is a label propagation step based on a more powerful similarity measure that however is only consistent within specific subsets of the data at a time.\r\nIn our biological setup, the information sources are the shape and the staining patterns of embryo images. We show\r\nexperimentally  that  while  neither  of  the  methods  can  be used by itself to achieve satisfactory results, their combina-\r\ntion achieves prediction quality comparable to human performance."}],"_id":"2294","citation":{"short":"T. Kazmar, E. Kvon, A. Stark, C. Lampert, in:, IEEE, 2013.","ieee":"T. Kazmar, E. Kvon, A. Stark, and C. Lampert, “Drosophila Embryo Stage Annotation using Label Propagation,” presented at the ICCV: International Conference on Computer Vision, Sydney, Australia, 2013.","apa":"Kazmar, T., Kvon, E., Stark, A., &#38; Lampert, C. (2013). Drosophila Embryo Stage Annotation using Label Propagation. Presented at the ICCV: International Conference on Computer Vision, Sydney, Australia: IEEE. <a href=\"https://doi.org/10.1109/ICCV.2013.139\">https://doi.org/10.1109/ICCV.2013.139</a>","mla":"Kazmar, Tomas, et al. <i>Drosophila Embryo Stage Annotation Using Label Propagation</i>. IEEE, 2013, doi:<a href=\"https://doi.org/10.1109/ICCV.2013.139\">10.1109/ICCV.2013.139</a>.","ista":"Kazmar T, Kvon E, Stark A, Lampert C. 2013. Drosophila Embryo Stage Annotation using Label Propagation. ICCV: International Conference on Computer Vision.","chicago":"Kazmar, Tomas, Evgeny Kvon, Alexander Stark, and Christoph Lampert. “Drosophila Embryo Stage Annotation Using Label Propagation.” IEEE, 2013. <a href=\"https://doi.org/10.1109/ICCV.2013.139\">https://doi.org/10.1109/ICCV.2013.139</a>.","ama":"Kazmar T, Kvon E, Stark A, Lampert C. Drosophila Embryo Stage Annotation using Label Propagation. In: IEEE; 2013. doi:<a href=\"https://doi.org/10.1109/ICCV.2013.139\">10.1109/ICCV.2013.139</a>"},"date_created":"2018-12-11T11:56:49Z","type":"conference","user_id":"317138e5-6ab7-11ef-aa6d-ffef3953e345","title":"Drosophila Embryo Stage Annotation using Label Propagation","day":"01","scopus_import":"1","isi":1},{"language":[{"iso":"eng"}],"author":[{"full_name":"Lampert, Christoph","orcid":"0000-0001-8622-7887","last_name":"Lampert","first_name":"Christoph","id":"40C20FD2-F248-11E8-B48F-1D18A9856A87"},{"last_name":"Nickisch","first_name":"Hannes","full_name":"Nickisch, Hannes"},{"full_name":"Harmeling, Stefan","last_name":"Harmeling","first_name":"Stefan"}],"publication_status":"published","month":"07","intvolume":"        36","publist_id":"4385","doi":"10.1109/TPAMI.2013.140","article_processing_charge":"No","fulldoi":"https://doi.org/10.1109/TPAMI.2013.140","publisher":"IEEE","publication":"IEEE Transactions on Pattern Analysis and Machine Intelligence","oa_version":"None","date_updated":"2025-09-29T14:09:57Z","quality_controlled":"1","volume":36,"external_id":{"isi":["000331450100005"]},"page":"453 - 465","year":"2013","_id":"2516","abstract":[{"text":"We study the problem of object recognition for categories for which we have no training examples, a task also called zero-data or zero-shot learning. This situation has hardly been studied in computer vision research, even though it occurs frequently: the world contains tens of thousands of different object classes and for only few of them image collections have been formed and suitably annotated. To tackle the problem we introduce attribute-based classification: objects are identified based on a high-level description that is phrased in terms of semantic attributes, such as the object's color or shape. Because the identification of each such property transcends the specific learning task at hand, the attribute classifiers can be pre-learned independently, e.g. from existing image datasets unrelated to the current task. Afterwards, new classes can be detected based on their attribute representation, without the need for a new training phase. In this paper we also introduce a new dataset, Animals with Attributes, of over 30,000 images of 50 animal classes, annotated with 85 semantic attributes. Extensive experiments on this and two more datasets show that attribute-based classification indeed is able to categorize images without access to any training images of the target classes.","lang":"eng"}],"department":[{"_id":"ChLa"}],"date_published":"2013-07-30T00:00:00Z","status":"public","day":"30","title":"Attribute-based classification for zero-shot learning of object categories","user_id":"317138e5-6ab7-11ef-aa6d-ffef3953e345","isi":1,"scopus_import":"1","issue":"3","date_created":"2018-12-11T11:58:08Z","citation":{"short":"C. Lampert, H. Nickisch, S. Harmeling, IEEE Transactions on Pattern Analysis and Machine Intelligence 36 (2013) 453–465.","ieee":"C. Lampert, H. Nickisch, and S. Harmeling, “Attribute-based classification for zero-shot learning of object categories,” <i>IEEE Transactions on Pattern Analysis and Machine Intelligence</i>, vol. 36, no. 3. IEEE, pp. 453–465, 2013.","apa":"Lampert, C., Nickisch, H., &#38; Harmeling, S. (2013). Attribute-based classification for zero-shot learning of object categories. <i>IEEE Transactions on Pattern Analysis and Machine Intelligence</i>. IEEE. <a href=\"https://doi.org/10.1109/TPAMI.2013.140\">https://doi.org/10.1109/TPAMI.2013.140</a>","mla":"Lampert, Christoph, et al. “Attribute-Based Classification for Zero-Shot Learning of Object Categories.” <i>IEEE Transactions on Pattern Analysis and Machine Intelligence</i>, vol. 36, no. 3, IEEE, 2013, pp. 453–65, doi:<a href=\"https://doi.org/10.1109/TPAMI.2013.140\">10.1109/TPAMI.2013.140</a>.","ama":"Lampert C, Nickisch H, Harmeling S. Attribute-based classification for zero-shot learning of object categories. <i>IEEE Transactions on Pattern Analysis and Machine Intelligence</i>. 2013;36(3):453-465. doi:<a href=\"https://doi.org/10.1109/TPAMI.2013.140\">10.1109/TPAMI.2013.140</a>","chicago":"Lampert, Christoph, Hannes Nickisch, and Stefan Harmeling. “Attribute-Based Classification for Zero-Shot Learning of Object Categories.” <i>IEEE Transactions on Pattern Analysis and Machine Intelligence</i>. IEEE, 2013. <a href=\"https://doi.org/10.1109/TPAMI.2013.140\">https://doi.org/10.1109/TPAMI.2013.140</a>.","ista":"Lampert C, Nickisch H, Harmeling S. 2013. Attribute-based classification for zero-shot learning of object categories. IEEE Transactions on Pattern Analysis and Machine Intelligence. 36(3), 453–465."},"type":"journal_article"},{"title":"The supervised IBP: Neighbourhood preserving infinite latent feature models","day":"11","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","scopus_import":1,"citation":{"ista":"Quadrianto N, Sharmanska V, Knowles D, Ghahramani Z. 2013. The supervised IBP: Neighbourhood preserving infinite latent feature models. Proceedings of the 29th conference uncertainty in Artificial Intelligence. UAI: Uncertainty in Artificial Intelligence, 527–536.","ama":"Quadrianto N, Sharmanska V, Knowles D, Ghahramani Z. The supervised IBP: Neighbourhood preserving infinite latent feature models. In: <i>Proceedings of the 29th Conference Uncertainty in Artificial Intelligence</i>. AUAI Press; 2013:527-536.","chicago":"Quadrianto, Novi, Viktoriia Sharmanska, David Knowles, and Zoubin Ghahramani. “The Supervised IBP: Neighbourhood Preserving Infinite Latent Feature Models.” In <i>Proceedings of the 29th Conference Uncertainty in Artificial Intelligence</i>, 527–36. AUAI Press, 2013.","short":"N. Quadrianto, V. Sharmanska, D. Knowles, Z. Ghahramani, in:, Proceedings of the 29th Conference Uncertainty in Artificial Intelligence, AUAI Press, 2013, pp. 527–536.","apa":"Quadrianto, N., Sharmanska, V., Knowles, D., &#38; Ghahramani, Z. (2013). The supervised IBP: Neighbourhood preserving infinite latent feature models. In <i>Proceedings of the 29th conference uncertainty in Artificial Intelligence</i> (pp. 527–536). Bellevue, WA, United States: AUAI Press.","ieee":"N. Quadrianto, V. Sharmanska, D. Knowles, and Z. Ghahramani, “The supervised IBP: Neighbourhood preserving infinite latent feature models,” in <i>Proceedings of the 29th conference uncertainty in Artificial Intelligence</i>, Bellevue, WA, United States, 2013, pp. 527–536.","mla":"Quadrianto, Novi, et al. “The Supervised IBP: Neighbourhood Preserving Infinite Latent Feature Models.” <i>Proceedings of the 29th Conference Uncertainty in Artificial Intelligence</i>, AUAI Press, 2013, pp. 527–36."},"date_created":"2018-12-11T11:58:09Z","type":"conference","_id":"2520","abstract":[{"lang":"eng","text":"We propose a probabilistic model to infer supervised latent variables in\r\nthe Hamming space from observed data. Our model allows simultaneous\r\ninference of the number of binary latent variables, and their values. The\r\nlatent variables preserve neighbourhood structure of the data in a sense\r\nthat objects in the same semantic concept have similar latent values, and\r\nobjects in different concepts have dissimilar latent values. We formulate\r\nthe supervised infinite latent variable problem based on an intuitive\r\nprinciple of pulling objects together if they are of the same type, and\r\npushing them apart if they are not. We then combine this principle with a\r\nflexible Indian Buffet Process prior on the latent variables. We show that\r\nthe inferred supervised latent variables can be directly used to perform a\r\nnearest neighbour search for the purpose of retrieval.  We introduce a new\r\napplication of dynamically extending hash codes, and show how to\r\neffectively couple the structure of the hash codes with continuously\r\ngrowing structure of the neighbourhood preserving infinite latent feature\r\nspace."}],"has_accepted_license":"1","file_date_updated":"2020-07-14T12:45:42Z","department":[{"_id":"ChLa"}],"date_published":"2013-07-11T00:00:00Z","status":"public","publisher":"AUAI Press","publication":"Proceedings of the 29th conference uncertainty in Artificial Intelligence","date_updated":"2023-02-23T10:46:36Z","quality_controlled":"1","oa_version":"Submitted Version","page":"527 - 536","file":[{"date_updated":"2020-07-14T12:45:42Z","file_id":"5134","checksum":"325f20c4b926bd74d39006b97df572bd","file_size":1117100,"file_name":"IST-2013-137-v1+1_QuaShaKnoGha13.pdf","creator":"system","relation":"main_file","content_type":"application/pdf","access_level":"open_access","date_created":"2018-12-12T10:15:16Z"}],"year":"2013","language":[{"iso":"eng"}],"author":[{"last_name":"Quadrianto","first_name":"Novi","full_name":"Quadrianto, Novi"},{"full_name":"Sharmanska, Viktoriia","orcid":"0000-0003-0192-9308","first_name":"Viktoriia","last_name":"Sharmanska","id":"2EA6D09E-F248-11E8-B48F-1D18A9856A87"},{"first_name":"David","last_name":"Knowles","full_name":"Knowles, David"},{"full_name":"Ghahramani, Zoubin","last_name":"Ghahramani","first_name":"Zoubin"}],"ddc":["000"],"month":"07","publication_status":"published","publist_id":"4381","conference":{"location":"Bellevue, WA, United States","start_date":"2013-07-11","name":"UAI: Uncertainty in Artificial Intelligence","end_date":"2013-07-15"},"pubrep_id":"137","oa":1,"publication_identifier":{"isbn":["9780974903996"]}},{"day":"04","title":"Beyond dataset bias: Multi-task unaligned shared knowledge transfer","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","scopus_import":1,"series_title":"Lecture Notes in Computer Science","date_created":"2018-12-11T12:00:30Z","citation":{"apa":"Tommasi, T., Quadrianto, N., Caputo, B., &#38; Lampert, C. (2013). Beyond dataset bias: Multi-task unaligned shared knowledge transfer. Presented at the ACCV: Asian Conference on Computer Vision, Daejeon, Korea: Springer. <a href=\"https://doi.org/10.1007/978-3-642-37331-2_1\">https://doi.org/10.1007/978-3-642-37331-2_1</a>","short":"T. Tommasi, N. Quadrianto, B. Caputo, C. Lampert, 7724 (2013) 1–15.","ieee":"T. Tommasi, N. Quadrianto, B. Caputo, and C. Lampert, “Beyond dataset bias: Multi-task unaligned shared knowledge transfer,” vol. 7724. Springer, pp. 1–15, 2013.","mla":"Tommasi, Tatiana, et al. <i>Beyond Dataset Bias: Multi-Task Unaligned Shared Knowledge Transfer</i>. Vol. 7724, Springer, 2013, pp. 1–15, doi:<a href=\"https://doi.org/10.1007/978-3-642-37331-2_1\">10.1007/978-3-642-37331-2_1</a>.","ama":"Tommasi T, Quadrianto N, Caputo B, Lampert C. Beyond dataset bias: Multi-task unaligned shared knowledge transfer. 2013;7724:1-15. doi:<a href=\"https://doi.org/10.1007/978-3-642-37331-2_1\">10.1007/978-3-642-37331-2_1</a>","chicago":"Tommasi, Tatiana, Novi Quadrianto, Barbara Caputo, and Christoph Lampert. “Beyond Dataset Bias: Multi-Task Unaligned Shared Knowledge Transfer.” Lecture Notes in Computer Science. Springer, 2013. <a href=\"https://doi.org/10.1007/978-3-642-37331-2_1\">https://doi.org/10.1007/978-3-642-37331-2_1</a>.","ista":"Tommasi T, Quadrianto N, Caputo B, Lampert C. 2013. Beyond dataset bias: Multi-task unaligned shared knowledge transfer. 7724, 1–15."},"type":"conference","_id":"2948","abstract":[{"text":"Many visual datasets are traditionally used to analyze the performance of different learning techniques. The evaluation is usually done within each dataset, therefore it is questionable if such results are a reliable indicator of true generalization ability. We propose here an algorithm to exploit the existing data resources when learning on a new multiclass problem. Our main idea is to identify an image representation that decomposes orthogonally into two subspaces: a part specific to each dataset, and a part generic to, and therefore shared between, all the considered source sets. This allows us to use the generic representation as un-biased reference knowledge for a novel classification task. By casting the method in the multi-view setting, we also make it possible to use different features for different databases. We call the algorithm MUST, Multitask Unaligned Shared knowledge Transfer. Through extensive experiments on five public datasets, we show that MUST consistently improves the cross-datasets generalization performance.","lang":"eng"}],"file_date_updated":"2020-07-14T12:45:55Z","has_accepted_license":"1","acknowledgement":"This work was supported by the PASCAL 2 Network of Excellence (TT) and by the Newton International Fellowship (NQ)","department":[{"_id":"ChLa"}],"date_published":"2013-04-04T00:00:00Z","status":"public","publisher":"Springer","oa_version":"Submitted Version","quality_controlled":"1","date_updated":"2020-08-11T10:09:54Z","page":"1 - 15","file":[{"file_size":1513620,"file_id":"5874","checksum":"a0a7234a89e2192af655b0d0ae3bf445","date_updated":"2020-07-14T12:45:55Z","date_created":"2019-01-22T14:03:11Z","access_level":"open_access","creator":"dernst","content_type":"application/pdf","relation":"main_file","file_name":"2012_ACCV_Tommasi.pdf"}],"volume":7724,"year":"2013","language":[{"iso":"eng"}],"author":[{"last_name":"Tommasi","first_name":"Tatiana","full_name":"Tommasi, Tatiana"},{"full_name":"Quadrianto, Novi","first_name":"Novi","last_name":"Quadrianto"},{"full_name":"Caputo, Barbara","last_name":"Caputo","first_name":"Barbara"},{"full_name":"Lampert, Christoph","id":"40C20FD2-F248-11E8-B48F-1D18A9856A87","first_name":"Christoph","last_name":"Lampert","orcid":"0000-0001-8622-7887"}],"month":"04","intvolume":"      7724","ddc":["000"],"publication_status":"published","conference":{"end_date":"2012-11-09","name":"ACCV: Asian Conference on Computer Vision","start_date":"2012-11-05","location":"Daejeon, Korea"},"publist_id":"3784","doi":"10.1007/978-3-642-37331-2_1","oa":1,"alternative_title":["LNCS"],"fulldoi":"https://doi.org/10.1007/978-3-642-37331-2_1"},{"type":"encyclopedia_article","year":"2013","citation":{"ama":"Quadrianto N, Lampert C. Kernel based learning. In: Dubitzky W, Wolkenhauer O, Cho K, Yokota H, eds. <i>Encyclopedia of Systems Biology</i>. Vol 3. Springer; 2013:1069-1069. doi:<a href=\"https://doi.org/10.1007/978-1-4419-9863-7_604\">10.1007/978-1-4419-9863-7_604</a>","chicago":"Quadrianto, Novi, and Christoph Lampert. “Kernel Based Learning.” In <i>Encyclopedia of Systems Biology</i>, edited by Werner Dubitzky, Olaf Wolkenhauer, Kwang Cho, and Hiroki Yokota, 3:1069–1069. Springer, 2013. <a href=\"https://doi.org/10.1007/978-1-4419-9863-7_604\">https://doi.org/10.1007/978-1-4419-9863-7_604</a>.","ista":"Quadrianto N, Lampert C. 2013.Kernel based learning. In: Encyclopedia of Systems Biology. vol. 3, 1069–1069.","mla":"Quadrianto, Novi, and Christoph Lampert. “Kernel Based Learning.” <i>Encyclopedia of Systems Biology</i>, edited by Werner Dubitzky et al., vol. 3, Springer, 2013, pp. 1069–1069, doi:<a href=\"https://doi.org/10.1007/978-1-4419-9863-7_604\">10.1007/978-1-4419-9863-7_604</a>.","short":"N. Quadrianto, C. Lampert, in:, W. Dubitzky, O. Wolkenhauer, K. Cho, H. Yokota (Eds.), Encyclopedia of Systems Biology, Springer, 2013, pp. 1069–1069.","ieee":"N. Quadrianto and C. Lampert, “Kernel based learning,” in <i>Encyclopedia of Systems Biology</i>, vol. 3, W. Dubitzky, O. Wolkenhauer, K. Cho, and H. Yokota, Eds. Springer, 2013, pp. 1069–1069.","apa":"Quadrianto, N., &#38; Lampert, C. (2013). Kernel based learning. In W. Dubitzky, O. Wolkenhauer, K. Cho, &#38; H. Yokota (Eds.), <i>Encyclopedia of Systems Biology</i> (Vol. 3, pp. 1069–1069). Springer. <a href=\"https://doi.org/10.1007/978-1-4419-9863-7_604\">https://doi.org/10.1007/978-1-4419-9863-7_604</a>"},"date_created":"2018-12-11T12:02:39Z","volume":3,"page":"1069 - 1069","quality_controlled":"1","date_updated":"2021-01-12T07:42:38Z","oa_version":"None","editor":[{"first_name":"Werner","last_name":"Dubitzky","full_name":"Dubitzky, Werner"},{"full_name":"Wolkenhauer, Olaf","first_name":"Olaf","last_name":"Wolkenhauer"},{"first_name":"Kwang","last_name":"Cho","full_name":"Cho, Kwang"},{"last_name":"Yokota","first_name":"Hiroki","full_name":"Yokota, Hiroki"}],"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","title":"Kernel based learning","publication":"Encyclopedia of Systems Biology","publisher":"Springer","day":"01","fulldoi":"https://doi.org/10.1007/978-1-4419-9863-7_604","status":"public","date_published":"2013-01-01T00:00:00Z","doi":"10.1007/978-1-4419-9863-7_604","department":[{"_id":"ChLa"}],"publist_id":"3314","intvolume":"         3","month":"01","publication_status":"published","author":[{"first_name":"Novi","last_name":"Quadrianto","full_name":"Quadrianto, Novi"},{"orcid":"0000-0001-8622-7887","last_name":"Lampert","first_name":"Christoph","id":"40C20FD2-F248-11E8-B48F-1D18A9856A87","full_name":"Lampert, Christoph"}],"_id":"3321","language":[{"iso":"eng"}]}]