{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/trusting-svm-for-piecewise-linear-cnns","title":"Trusting SVM for Piecewise Linear CNNs","arxiv_id":"1611.02185","date":"2016-11-07","proceeding":null,"authors":["Leonard Berrada","Andrew Zisserman","M. Pawan Kumar"],"abstract":"We present a novel layerwise optimization algorithm for the learning\nobjective of Piecewise-Linear Convolutional Neural Networks (PL-CNNs), a large\nclass of convolutional neural networks. Specifically, PL-CNNs employ piecewise\nlinear non-linearities such as the commonly used ReLU and max-pool, and an SVM\nclassifier as the final layer. The key observation of our approach is that the\nproblem corresponding to the parameter estimation of a layer can be formulated\nas a difference-of-convex (DC) program, which happens to be a latent structured\nSVM. We optimize the DC program using the concave-convex procedure, which\nrequires us to iteratively solve a structured SVM problem. This allows to\ndesign an optimization algorithm with an optimal learning rate that does not\nrequire any tuning. Using the MNIST, CIFAR and ImageNet data sets, we show that\nour approach always improves over the state of the art variants of\nbackpropagation and scales to large data and large network settings.","url_abs":"http://arxiv.org/abs/1611.02185v5","url_pdf":"http://arxiv.org/pdf/1611.02185v5.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"trusting-svm-for-piecewise-linear-cnns","repo_url":"https://github.com/oval-group/pl-cnn","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"trusting-svm-for-piecewise-linear-cnns","repo_url":"https://github.com/lukasruff/Deep-SVDD","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"parameter-estimation","task_name":"parameter estimation"}],"methods":[{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"svm","method_name":"SVM"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}