{"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/dynamic-channel-pruning-feature-boosting-and","title":"Dynamic Channel Pruning: Feature Boosting and Suppression","arxiv_id":"1810.05331","date":"2018-10-12","proceeding":"ICLR 2019 5","authors":["Xitong Gao","Yiren Zhao","Łukasz Dudziak","Robert Mullins","Cheng-Zhong Xu"],"abstract":"Making deep convolutional neural networks more accurate typically comes at\nthe cost of increased computational and memory resources. In this paper, we\nreduce this cost by exploiting the fact that the importance of features\ncomputed by convolutional layers is highly input-dependent, and propose feature\nboosting and suppression (FBS), a new method to predictively amplify salient\nconvolutional channels and skip unimportant ones at run-time. FBS introduces\nsmall auxiliary connections to existing convolutional layers. In contrast to\nchannel pruning methods which permanently remove channels, it preserves the\nfull network structures and accelerates convolution by dynamically skipping\nunimportant input and output channels. FBS-augmented networks are trained with\nconventional stochastic gradient descent, making it readily available for many\nstate-of-the-art CNNs. We compare FBS to a range of existing channel pruning\nand dynamic execution schemes and demonstrate large improvements on ImageNet\nclassification. Experiments show that FBS can respectively provide $5\\times$\nand $2\\times$ savings in compute on VGG-16 and ResNet-18, both with less than\n$0.6\\%$ top-5 accuracy loss.","url_abs":"http://arxiv.org/abs/1810.05331v2","url_pdf":"http://arxiv.org/pdf/1810.05331v2.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":"dynamic-channel-pruning-feature-boosting-and","repo_url":"https://github.com/deep-fry/mayo","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"dynamic-channel-pruning-feature-boosting-and","repo_url":"https://github.com/YOUSIKI/PyTorch-FBS","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"model-compression","task_name":"Model Compression"},{"task_slug":"network-pruning","task_name":"Network Pruning"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"pruning","method_name":"Pruning"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1810.05331","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"1810.05331"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/YOUSIKI/PyTorch-FBS","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/deep-fry/mayo","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"ran_draft_wrong":2,"unverified":5},"by_repo_kind":{"official":{"samples":3,"ran":0,"repositories":1},"listed":{"samples":4,"ran":2,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"c61b8516007e275a","entry":"get_in_channels","repo":"YOUSIKI/PyTorch-FBS","repo_kind":"listed","path":"models/custom.py","file_url":"https://github.com/YOUSIKI/PyTorch-FBS/blob/HEAD/models/custom.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"well_formed","behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"c61b8516007e275a"}},{"code_sha256_prefix":"8b74cd9e7bbcae25","entry":"get_out_channels","repo":"YOUSIKI/PyTorch-FBS","repo_kind":"listed","path":"models/custom.py","file_url":"https://github.com/YOUSIKI/PyTorch-FBS/blob/HEAD/models/custom.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"well_formed","behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"8b74cd9e7bbcae25"}},{"code_sha256_prefix":"350baadaa22ae4dd","entry":"conv1x1","repo":"YOUSIKI/PyTorch-FBS","repo_kind":"listed","path":"models/resnet.py","file_url":"https://github.com/YOUSIKI/PyTorch-FBS/blob/HEAD/models/resnet.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"350baadaa22ae4dd"}},{"code_sha256_prefix":"aa53dd1441e0a7d8","entry":"conv3x3","repo":"YOUSIKI/PyTorch-FBS","repo_kind":"listed","path":"models/resnet.py","file_url":"https://github.com/YOUSIKI/PyTorch-FBS/blob/HEAD/models/resnet.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"aa53dd1441e0a7d8"}},{"code_sha256_prefix":"3088ca7dabe32a62","entry":"mask_join","repo":"deep-fry/mayo","repo_kind":"official","path":"mayo/net/tf/estimate.py","file_url":"https://github.com/deep-fry/mayo/blob/HEAD/mayo/net/tf/estimate.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"3088ca7dabe32a62"}},{"code_sha256_prefix":"0667f26e9f8bd466","entry":"multiply","repo":"deep-fry/mayo","repo_kind":"official","path":"mayo/net/tf/estimate.py","file_url":"https://github.com/deep-fry/mayo/blob/HEAD/mayo/net/tf/estimate.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"0667f26e9f8bd466"}},{"code_sha256_prefix":"9a0f792b34207b94","entry":"use_name_not_scope","repo":"deep-fry/mayo","repo_kind":"official","path":"mayo/net/tf/transform.py","file_url":"https://github.com/deep-fry/mayo/blob/HEAD/mayo/net/tf/transform.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"9a0f792b34207b94"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}