Deadwooding: Robust Global Pruning for Deep Neural Networks

Deadwooding: Robust Global Pruning for Deep Neural Networks. Kaur, S., Fioretto, F., & Salekin, A. September, 2022. arXiv:2202.05226 [cs]

Paper doi abstract bibtex

The ability of Deep Neural Networks to approximate highly complex functions is key to their success. This benefit, however, comes at the expense of a large model size, which challenges its deployment in resource-constrained environments. Pruning is an effective technique used to limit this issue, but often comes at the cost of reduced accuracy and adversarial robustness. This paper addresses these shortcomings and introduces Deadwooding, a novel global pruning technique that exploits a Lagrangian Dual method to encourage model sparsity while retaining accuracy and ensuring robustness. The resulting model is shown to significantly outperform the state-of-the-art studies in measures of robustness and accuracy.

@misc{kaur_deadwooding_2022,
	title = {Deadwooding: {Robust} {Global} {Pruning} for {Deep} {Neural} {Networks}},
	shorttitle = {Deadwooding},
	url = {http://arxiv.org/abs/2202.05226},
	doi = {10.48550/arXiv.2202.05226},
	abstract = {The ability of Deep Neural Networks to approximate highly complex functions is key to their success. This benefit, however, comes at the expense of a large model size, which challenges its deployment in resource-constrained environments. Pruning is an effective technique used to limit this issue, but often comes at the cost of reduced accuracy and adversarial robustness. This paper addresses these shortcomings and introduces Deadwooding, a novel global pruning technique that exploits a Lagrangian Dual method to encourage model sparsity while retaining accuracy and ensuring robustness. The resulting model is shown to significantly outperform the state-of-the-art studies in measures of robustness and accuracy.},
	urldate = {2023-01-06},
	publisher = {arXiv},
	author = {Kaur, Sawinder and Fioretto, Ferdinando and Salekin, Asif},
	month = sep,
	year = {2022},
	note = {arXiv:2202.05226 [cs]},
	keywords = {Computer Science - Machine Learning},
}

Downloads: 0

{"_id":"fjzmJpisX66qTGAKy","bibbaseid":"kaur-fioretto-salekin-deadwoodingrobustglobalpruningfordeepneuralnetworks-2022","author_short":["Kaur, S.","Fioretto, F.","Salekin, A."],"bibdata":{"bibtype":"misc","type":"misc","title":"Deadwooding: Robust Global Pruning for Deep Neural Networks","shorttitle":"Deadwooding","url":"http://arxiv.org/abs/2202.05226","doi":"10.48550/arXiv.2202.05226","abstract":"The ability of Deep Neural Networks to approximate highly complex functions is key to their success. This benefit, however, comes at the expense of a large model size, which challenges its deployment in resource-constrained environments. Pruning is an effective technique used to limit this issue, but often comes at the cost of reduced accuracy and adversarial robustness. This paper addresses these shortcomings and introduces Deadwooding, a novel global pruning technique that exploits a Lagrangian Dual method to encourage model sparsity while retaining accuracy and ensuring robustness. The resulting model is shown to significantly outperform the state-of-the-art studies in measures of robustness and accuracy.","urldate":"2023-01-06","publisher":"arXiv","author":[{"propositions":[],"lastnames":["Kaur"],"firstnames":["Sawinder"],"suffixes":[]},{"propositions":[],"lastnames":["Fioretto"],"firstnames":["Ferdinando"],"suffixes":[]},{"propositions":[],"lastnames":["Salekin"],"firstnames":["Asif"],"suffixes":[]}],"month":"September","year":"2022","note":"arXiv:2202.05226 [cs]","keywords":"Computer Science - Machine Learning","bibtex":"@misc{kaur_deadwooding_2022,\n\ttitle = {Deadwooding: {Robust} {Global} {Pruning} for {Deep} {Neural} {Networks}},\n\tshorttitle = {Deadwooding},\n\turl = {http://arxiv.org/abs/2202.05226},\n\tdoi = {10.48550/arXiv.2202.05226},\n\tabstract = {The ability of Deep Neural Networks to approximate highly complex functions is key to their success. This benefit, however, comes at the expense of a large model size, which challenges its deployment in resource-constrained environments. Pruning is an effective technique used to limit this issue, but often comes at the cost of reduced accuracy and adversarial robustness. This paper addresses these shortcomings and introduces Deadwooding, a novel global pruning technique that exploits a Lagrangian Dual method to encourage model sparsity while retaining accuracy and ensuring robustness. The resulting model is shown to significantly outperform the state-of-the-art studies in measures of robustness and accuracy.},\n\turldate = {2023-01-06},\n\tpublisher = {arXiv},\n\tauthor = {Kaur, Sawinder and Fioretto, Ferdinando and Salekin, Asif},\n\tmonth = sep,\n\tyear = {2022},\n\tnote = {arXiv:2202.05226 [cs]},\n\tkeywords = {Computer Science - Machine Learning},\n}\n\n","author_short":["Kaur, S.","Fioretto, F.","Salekin, A."],"key":"kaur_deadwooding_2022","id":"kaur_deadwooding_2022","bibbaseid":"kaur-fioretto-salekin-deadwoodingrobustglobalpruningfordeepneuralnetworks-2022","role":"author","urls":{"Paper":"http://arxiv.org/abs/2202.05226"},"keyword":["Computer Science - Machine Learning"],"metadata":{"authorlinks":{}},"html":""},"bibtype":"misc","biburl":"https://bibbase.org/zotero/mxmplx","dataSources":["aXmRAq63YsH7a3ufx"],"keywords":["computer science - machine learning"],"search_terms":["deadwooding","robust","global","pruning","deep","neural","networks","kaur","fioretto","salekin"],"title":"Deadwooding: Robust Global Pruning for Deep Neural Networks","year":2022}