{"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/scattering-networks-for-hybrid-representation","title":"Scattering Networks for Hybrid Representation Learning","arxiv_id":"1809.06367","date":"2018-09-17","proceeding":null,"authors":["Edouard Oyallon","Sergey Zagoruyko","Gabriel Huang","Nikos Komodakis","Simon Lacoste-Julien","Matthew Blaschko","Eugene Belilovsky"],"abstract":"Scattering networks are a class of designed Convolutional Neural Networks\n(CNNs) with fixed weights. We argue they can serve as generic representations\nfor modelling images. In particular, by working in scattering space, we achieve\ncompetitive results both for supervised and unsupervised learning tasks, while\nmaking progress towards constructing more interpretable CNNs. For supervised\nlearning, we demonstrate that the early layers of CNNs do not necessarily need\nto be learned, and can be replaced with a scattering network instead. Indeed,\nusing hybrid architectures, we achieve the best results with predefined\nrepresentations to-date, while being competitive with end-to-end learned CNNs.\nSpecifically, even applying a shallow cascade of small-windowed scattering\ncoefficients followed by 1$\\times$1-convolutions results in AlexNet accuracy on\nthe ILSVRC2012 classification task. Moreover, by combining scattering networks\nwith deep residual networks, we achieve a single-crop top-5 error of 11.4% on\nILSVRC2012. Also, we show they can yield excellent performance in the small\nsample regime on CIFAR-10 and STL-10 datasets, exceeding their end-to-end\ncounterparts, through their ability to incorporate geometrical priors. For\nunsupervised learning, scattering coefficients can be a competitive\nrepresentation that permits image recovery. We use this fact to train hybrid\nGANs to generate images. Finally, we empirically analyze several properties\nrelated to stability and reconstruction of images from scattering coefficients.","url_abs":"http://arxiv.org/abs/1809.06367v1","url_pdf":"http://arxiv.org/pdf/1809.06367v1.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":"scattering-networks-for-hybrid-representation","repo_url":"https://github.com/bentherien/ParametricScatteringNetworks","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"BSD-3-Clause"}}],"tasks":[{"task_slug":"representation-learning","task_name":"Representation Learning"}],"methods":[{"method_slug":"1x1-convolution","method_name":"1x1 Convolution"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"grouped-convolution","method_name":"Grouped Convolution"},{"method_slug":"local-response-normalization","method_name":"Local Response Normalization"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1809.06367","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}