{"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/log-densenet-how-to-sparsify-a-densenet","title":"Log-DenseNet: How to Sparsify a DenseNet","arxiv_id":"1711.00002","date":"2017-10-30","proceeding":"ICLR 2018 1","authors":["Hanzhang Hu","Debadeepta Dey","Allison Del Giorno","Martial Hebert","J. Andrew Bagnell"],"abstract":"Skip connections are increasingly utilized by deep neural networks to improve\naccuracy and cost-efficiency. In particular, the recent DenseNet is efficient\nin computation and parameters, and achieves state-of-the-art predictions by\ndirectly connecting each feature layer to all previous ones. However,\nDenseNet's extreme connectivity pattern may hinder its scalability to high\ndepths, and in applications like fully convolutional networks, full DenseNet\nconnections are prohibitively expensive. This work first experimentally shows\nthat one key advantage of skip connections is to have short distances among\nfeature layers during backpropagation. Specifically, using a fixed number of\nskip connections, the connection patterns with shorter backpropagation distance\namong layers have more accurate predictions. Following this insight, we propose\na connection template, Log-DenseNet, which, in comparison to DenseNet, only\nslightly increases the backpropagation distances among layers from 1 to ($1 +\n\\log_2 L$), but uses only $L\\log_2 L$ total connections instead of $O(L^2)$.\nHence, Log-DenseNets are easier than DenseNets to implement and to scale. We\ndemonstrate the effectiveness of our design principle by showing better\nperformance than DenseNets on tabula rasa semantic segmentation, and\ncompetitive results on visual recognition.","url_abs":"http://arxiv.org/abs/1711.00002v1","url_pdf":"http://arxiv.org/pdf/1711.00002v1.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":"log-densenet-how-to-sparsify-a-densenet","repo_url":"https://github.com/agassi4013/Log-DenseNet-Tensorflow","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}}],"tasks":[{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"}],"methods":[{"method_slug":"1x1-convolution","method_name":"1x1 Convolution"},{"method_slug":"average-pooling","method_name":"Average Pooling"},{"method_slug":"batch-normalization","method_name":"Batch Normalization"},{"method_slug":"concatenated-skip-connection","method_name":"Concatenated Skip Connection"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"dense-block","method_name":"Dense Block"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"global-average-pooling","method_name":"Global Average Pooling"},{"method_slug":"kaiming-initialization","method_name":"Kaiming Initialization"},{"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":{"syntology_url":"https://syntology.ai/paper/1711.00002","atlas_url":"https://app.syntology.ai/?focus=1711.00002","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}