{"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/beam-search-for-learning-a-deep-convolutional","title":"Beam Search for Learning a Deep Convolutional Neural Network of 3D Shapes","arxiv_id":"1612.04774","date":"2016-12-14","proceeding":null,"authors":["Xu Xu","Sinisa Todorovic"],"abstract":"This paper addresses 3D shape recognition. Recent work typically represents a\n3D shape as a set of binary variables corresponding to 3D voxels of a uniform\n3D grid centered on the shape, and resorts to deep convolutional neural\nnetworks(CNNs) for modeling these binary variables. Robust learning of such\nCNNs is currently limited by the small datasets of 3D shapes available, an\norder of magnitude smaller than other common datasets in computer vision.\nRelated work typically deals with the small training datasets using a number of\nad hoc, hand-tuning strategies. To address this issue, we formulate CNN\nlearning as a beam search aimed at identifying an optimal CNN architecture,\nnamely, the number of layers, nodes, and their connectivity in the network, as\nwell as estimating parameters of such an optimal CNN. Each state of the beam\nsearch corresponds to a candidate CNN. Two types of actions are defined to add\nnew convolutional filters or new convolutional layers to a parent CNN, and thus\ntransition to children states. The utility function of each action is\nefficiently computed by transferring parameter values of the parent CNN to its\nchildren, thereby enabling an efficient beam search. Our experimental\nevaluation on the 3D ModelNet dataset demonstrates that our model pursuit using\nthe beam search yields a CNN with superior performance on 3D shape\nclassification than the state of the art.","url_abs":"http://arxiv.org/abs/1612.04774v1","url_pdf":"http://arxiv.org/pdf/1612.04774v1.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":"beam-search-for-learning-a-deep-convolutional","repo_url":"https://github.com/xuxucmkox/3D-shape-Classification-Beam-Search","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"3d-shape-retrieval","task_name":"3D Shape Classification"},{"task_slug":"3d-shape-recognition","task_name":"3D Shape Recognition"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}