{"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/studying-the-plasticity-in-deep-convolutional","title":"Studying the Plasticity in Deep Convolutional Neural Networks using Random Pruning","arxiv_id":"1812.10240","date":"2018-12-26","proceeding":null,"authors":["Deepak Mittal","Shweta Bhardwaj","Mitesh M. Khapra","Balaraman Ravindran"],"abstract":"Recently there has been a lot of work on pruning filters from deep\nconvolutional neural networks (CNNs) with the intention of reducing\ncomputations.The key idea is to rank the filters based on a certain criterion\n(say, l1-norm) and retain only the top ranked filters. Once the low scoring\nfilters are pruned away the remainder of the network is fine tuned and is shown\nto give performance comparable to the original unpruned network. In this work,\nwe report experiments which suggest that the comparable performance of the\npruned network is not due to the specific criterion chosen but due to the\ninherent plasticity of deep neural networks which allows them to recover from\nthe loss of pruned filters once the rest of the filters are fine-tuned.\nSpecifically we show counter-intuitive results wherein by randomly pruning\n25-50% filters from deep CNNs we are able to obtain the same performance as\nobtained by using state-of-the-art pruning methods. We empirically validate our\nclaims by doing an exhaustive evaluation with VGG-16 and ResNet-50. We also\nevaluate a real world scenario where a CNN trained on all 1000 ImageNet classes\nneeds to be tested on only a small set of classes at test time (say, only\nanimals). We create a new benchmark dataset from ImageNet to evaluate such\nclass specific pruning and show that even here a random pruning strategy gives\nclose to state-of-the-art performance. Unlike existing approaches which mainly\nfocus on the task of image classification, in this work we also report results\non object detection and image segmentation. We show that using a simple random\npruning strategy we can achieve significant speed up in object detection (74%\nimprovement in fps) while retaining the same accuracy as that of the original\nFaster RCNN model. Similarly we show that the performance of a pruned\nSegmentation Network (SegNet) is actually very similar to that of the original\nunpruned SegNet.","url_abs":"http://arxiv.org/abs/1812.10240v1","url_pdf":"http://arxiv.org/pdf/1812.10240v1.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":"studying-the-plasticity-in-deep-convolutional","repo_url":"https://github.com/marcoancona/TorchPruner","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"image-segmentation","task_name":"Image Segmentation"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"},{"task_slug":"image-classification","task_name":"image-classification"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[{"method_slug":"batch-normalization","method_name":"Batch Normalization"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"kaiming-initialization","method_name":"Kaiming Initialization"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"pruning","method_name":"Pruning"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"speed","method_name":"SPEED"},{"method_slug":"segnet","method_name":"SegNet"},{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1812.10240","atlas_url":"https://app.syntology.ai/?focus=1812.10240","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}