{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,11,12]],"date-time":"2026-11-12T02:14:05Z","timestamp":1794449645712,"version":"4.9.1"},"reference-count":35,"publisher":"Association for Computing Machinery (ACM)","issue":"1-4","license":[{"start":{"date-parts":[[2019,11,30]],"date-time":"2019-11-30T00:00:00Z","timestamp":1575072000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/www.acm.org\/publications\/policies\/copyright_policy#Background"}],"content-domain":{"domain":["dl.acm.org"],"crossmark-restriction":true},"short-container-title":["ACM Trans. Comput. Syst."],"published-print":{"date-parts":[[2019,11,30]]},"abstract":"<jats:p>\n                    We present a novel method for neural network quantization. Our method, named\n                    <jats:italic>UNIQ<\/jats:italic>\n                    , emulates a non-uniform\n                    <jats:italic>k<\/jats:italic>\n                    -quantile quantizer and adapts the model to perform well with quantized weights by injecting noise to the weights at training time. As a by-product of injecting noise to weights, we find that activations can also be quantized to as low as 8-bit with only a minor accuracy degradation. Our non-uniform quantization approach provides a novel alternative to the existing uniform quantization techniques for neural networks. We further propose a novel complexity metric of number of bit operations performed (BOPs), and we show that this metric has a linear relation with logic utilization and power. We suggest evaluating the trade-off of accuracy vs. complexity (BOPs). The proposed method, when evaluated on ResNet18\/34\/50 and MobileNet on ImageNet, outperforms the prior state of the art both in the low-complexity regime and the high accuracy regime. We demonstrate the practical applicability of this approach, by implementing our non-uniformly quantized CNN on FPGA.\n                  <\/jats:p>","DOI":"10.1145\/3444943","type":"journal-article","created":{"date-parts":[[2021,3,26]],"date-time":"2021-03-26T12:31:19Z","timestamp":1616761879000},"page":"1-15","update-policy":"https:\/\/doi.org\/10.1145\/crossmark-policy","source":"Crossref","is-referenced-by-count":92,"title":["UNIQ"],"prefix":"10.1145","volume":"37","author":[{"given":"Chaim","family":"Baskin","sequence":"first","affiliation":[{"name":"Technion, Technion, Haifa"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Natan","family":"Liss","sequence":"additional","affiliation":[{"name":"Technion, Technion, Haifa"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Eli","family":"Schwartz","sequence":"additional","affiliation":[{"name":"Tel Aviv University, Tel Aviv, Israel"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5400-9321","authenticated-orcid":false,"given":"Evgenii","family":"Zheltonozhskii","sequence":"additional","affiliation":[{"name":"Technion, Technion, Haifa"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Raja","family":"Giryes","sequence":"additional","affiliation":[{"name":"Tel Aviv University, Tel Aviv, Israel"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Alex M.","family":"Bronstein","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Avi","family":"Mendelson","sequence":"additional","affiliation":[{"name":"Technion, Technion, Haifa"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"320","published-online":{"date-parts":[[2021,3,26]]},"reference":[{"key":"e_1_2_1_1_1","volume-title":"Proceedings of the International Conference on Learning Representations (ICLR\u201918)","author":"Alexander","unstructured":"Alexander G. Anderson and Cory P. Berg. 2018. The high-dimensional geometry of binary neural networks . In Proceedings of the International Conference on Learning Representations (ICLR\u201918) . Alexander G. Anderson and Cory P. Berg. 2018. The high-dimensional geometry of binary neural networks. In Proceedings of the International Conference on Learning Representations (ICLR\u201918)."},{"key":"e_1_2_1_2_1","volume-title":"Proceedings of the 32nd International Conference on Machine Learning (PMLR\u201915)","author":"Blundell Charles","year":"2015","unstructured":"Charles Blundell , Julien Cornebise , Koray Kavukcuoglu , and Daan Wierstra . 2015 . Weight uncertainty in neural network . In Proceedings of the 32nd International Conference on Machine Learning (PMLR\u201915) . 1613--1622. Charles Blundell, Julien Cornebise, Koray Kavukcuoglu, and Daan Wierstra. 2015. Weight uncertainty in neural network. In Proceedings of the 32nd International Conference on Machine Learning (PMLR\u201915). 1613--1622."},{"key":"e_1_2_1_3_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.574"},{"key":"e_1_2_1_4_1","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2017.2699184"},{"key":"e_1_2_1_5_1","doi-asserted-by":"publisher","DOI":"10.1109\/JSSC.2017.2749425"},{"key":"e_1_2_1_6_1","doi-asserted-by":"publisher","DOI":"10.5244\/C.31.189"},{"key":"e_1_2_1_7_1","doi-asserted-by":"publisher","DOI":"10.1109\/18.720541"},{"key":"e_1_2_1_8_1","volume-title":"Proceedings of the 32nd International Conference on Machine Learning (ICML\u201915)","author":"Gupta Suyog","year":"2015","unstructured":"Suyog Gupta , Ankur Agrawal , Kailash Gopalakrishnan , and Pritish Narayanan . 2015 . Deep learning with limited numerical precision . In Proceedings of the 32nd International Conference on Machine Learning (ICML\u201915) . 1737--1746. Suyog Gupta, Ankur Agrawal, Kailash Gopalakrishnan, and Pritish Narayanan. 2015. Deep learning with limited numerical precision. In Proceedings of the 32nd International Conference on Machine Learning (ICML\u201915). 1737--1746."},{"key":"e_1_2_1_9_1","volume-title":"Proceedings of the International Conference on Learning Representations (ICLR\u201916)","author":"Han Song","unstructured":"Song Han , Huizi Mao , and William J. Dally . 2016. Deep compression: Compressing deep neural networks with pruning, trained quantization and Huffman coding . In Proceedings of the International Conference on Learning Representations (ICLR\u201916) . Song Han, Huizi Mao, and William J. Dally. 2016. Deep compression: Compressing deep neural networks with pruning, trained quantization and Huffman coding. In Proceedings of the International Conference on Learning Representations (ICLR\u201916)."},{"key":"e_1_2_1_10_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"e_1_2_1_11_1","doi-asserted-by":"publisher","DOI":"10.1109\/MSP.2012.2205597"},{"key":"e_1_2_1_12_1","unstructured":"Andrew G. Howard Menglong Zhu Bo Chen Dmitry Kalenichenko Weijun Wang Tobias Weyand Marco Andreetto and Hartwig Adam. 2017. MobileNets: Efficient convolutional neural networks for mobile vision applications. arXiv:1704.04861  Andrew G. Howard Menglong Zhu Bo Chen Dmitry Kalenichenko Weijun Wang Tobias Weyand Marco Andreetto and Hartwig Adam. 2017. MobileNets: Efficient convolutional neural networks for mobile vision applications. arXiv:1704.04861"},{"key":"e_1_2_1_13_1","unstructured":"Itay Hubara Matthieu Courbariaux Daniel Soudry Ran El-Yaniv and Yoshua Bengio. 2016. Binarized neural networks. In Advances in Neural Information Processing Systems. 4107--4115.  Itay Hubara Matthieu Courbariaux Daniel Soudry Ran El-Yaniv and Yoshua Bengio. 2016. Binarized neural networks. In Advances in Neural Information Processing Systems. 4107--4115."},{"key":"e_1_2_1_14_1","doi-asserted-by":"publisher","DOI":"10.5555\/3122009.3242044"},{"key":"e_1_2_1_15_1","doi-asserted-by":"publisher","DOI":"10.1109\/MICRO.2016.7783722"},{"key":"e_1_2_1_16_1","volume-title":"Learning Multiple Layers of Features From Tiny Images. Master\u2019s Thesis. Department of Computer Science","author":"Krizhevsky Alex","unstructured":"Alex Krizhevsky . 2009. Learning Multiple Layers of Features From Tiny Images. Master\u2019s Thesis. Department of Computer Science , University of Toronto . Alex Krizhevsky. 2009. Learning Multiple Layers of Features From Tiny Images. Master\u2019s Thesis. Department of Computer Science, University of Toronto."},{"key":"e_1_2_1_17_1","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v29i1.9513"},{"key":"e_1_2_1_18_1","volume-title":"Proceedings of the 2017 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP\u201917)","author":"Lee Edward H.","unstructured":"Edward H. Lee , Daisuke Miyashita , Elaina Chai , Boris Murmann , and S. Simon Wong . 2017. LogNet: Energy-efficient neural networks using logarithmic computation . In Proceedings of the 2017 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP\u201917) . IEEE, Los Alamitos, CA, 5900--5904. Edward H. Lee, Daisuke Miyashita, Elaina Chai, Boris Murmann, and S. Simon Wong. 2017. LogNet: Energy-efficient neural networks using logarithmic computation. In Proceedings of the 2017 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP\u201917). IEEE, Los Alamitos, CA, 5900--5904."},{"key":"e_1_2_1_19_1","doi-asserted-by":"publisher","DOI":"10.1109\/TIT.1982.1056489"},{"key":"e_1_2_1_20_1","first-page":"I","article-title":"Bayesian compression for deep learning","volume":"30","author":"Louizos Christos","year":"2017","unstructured":"Christos Louizos , Karen Ullrich , and Max Welling . 2017 . Bayesian compression for deep learning . In Advances in Neural Information Processing Systems 30 , I . Guyon, U. V. Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett (Eds.). Curran Associates, Red Hook, NY, 3288--3298. Christos Louizos, Karen Ullrich, and Max Welling. 2017. Bayesian compression for deep learning. In Advances in Neural Information Processing Systems 30, I. Guyon, U. V. Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett (Eds.). Curran Associates, Red Hook, NY, 3288--3298.","journal-title":"Advances in Neural Information Processing Systems"},{"key":"e_1_2_1_21_1","volume-title":"Proceedings of the International Conference on Learning Representations (ICLR\u201918)","author":"Mishra Asit","year":"2018","unstructured":"Asit Mishra and Debbie Marr . 2018 . Apprentice: Using knowledge distillation techniques to improve low-precision network accuracy . In Proceedings of the International Conference on Learning Representations (ICLR\u201918) . Asit Mishra and Debbie Marr. 2018. Apprentice: Using knowledge distillation techniques to improve low-precision network accuracy. In Proceedings of the International Conference on Learning Representations (ICLR\u201918)."},{"key":"e_1_2_1_22_1","volume-title":"Proceedings of the International Conference on Learning Representations (ICLR\u201918)","author":"Mishra Asit","year":"2018","unstructured":"Asit Mishra , Eriko Nurvitadhi , Jeffrey J. Cook , and Debbie Marr . 2018 . WRPN: Wide reduced-precision networks . In Proceedings of the International Conference on Learning Representations (ICLR\u201918) . Asit Mishra, Eriko Nurvitadhi, Jeffrey J. Cook, and Debbie Marr. 2018. WRPN: Wide reduced-precision networks. In Proceedings of the International Conference on Learning Representations (ICLR\u201918)."},{"key":"e_1_2_1_23_1","volume-title":"Proceedings of the International Conference on Machine Learning.","author":"Molchanov Dmitry","year":"2017","unstructured":"Dmitry Molchanov , Arsenii Ashukha , and Dmitry Vetrov . 2017 . Variational dropout sparsifies deep neural networks . In Proceedings of the International Conference on Machine Learning. Dmitry Molchanov, Arsenii Ashukha, and Dmitry Vetrov. 2017. Variational dropout sparsifies deep neural networks. In Proceedings of the International Conference on Machine Learning."},{"key":"e_1_2_1_24_1","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-01225-0_36"},{"key":"e_1_2_1_25_1","volume-title":"Proceedings of the International Conference on Learning Representations (ICLR\u201918)","author":"Polino Antonio","year":"2018","unstructured":"Antonio Polino , Razvan Pascanu , and Dan Alistarh . 2018 . Model compression via distillation and quantization . In Proceedings of the International Conference on Learning Representations (ICLR\u201918) . Antonio Polino, Razvan Pascanu, and Dan Alistarh. 2018. Model compression via distillation and quantization. In Proceedings of the International Conference on Learning Representations (ICLR\u201918)."},{"key":"e_1_2_1_26_1","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-46493-0_32"},{"key":"e_1_2_1_27_1","doi-asserted-by":"publisher","DOI":"10.1007\/s11263-015-0816-y"},{"key":"e_1_2_1_28_1","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA.2018.00069"},{"key":"e_1_2_1_29_1","doi-asserted-by":"publisher","DOI":"10.1109\/EMC2.2018.00011"},{"key":"e_1_2_1_30_1","volume-title":"Proceedings of the International Conference on Learning Representations (ICLR\u201917)","author":"Ullrich Karen","year":"2017","unstructured":"Karen Ullrich , Edward Meeds , and Max Welling . 2017 . Soft weight-sharing for neural network compression . In Proceedings of the International Conference on Learning Representations (ICLR\u201917) . Karen Ullrich, Edward Meeds, and Max Welling. 2017. Soft weight-sharing for neural network compression. In Proceedings of the International Conference on Learning Representations (ICLR\u201917)."},{"key":"e_1_2_1_31_1","unstructured":"Yuhui Xu Yongzhuang Wang Aojun Zhou Weiyao Lin and Hongkai Xiong. 2018. Deep neural network compression with single and multiple level quantization. arXiv:1803.03289  Yuhui Xu Yongzhuang Wang Aojun Zhou Weiyao Lin and Hongkai Xiong. 2018. Deep neural network compression with single and multiple level quantization. arXiv:1803.03289"},{"key":"e_1_2_1_32_1","volume-title":"Proceedings of the International Conference on Learning Representations (ICLR\u201917)","author":"Zhou Aojun","year":"2017","unstructured":"Aojun Zhou , Anbang Yao , Yiwen Guo , Lin Xu , and Yurong Chen . 2017 . Incremental network quantization: Towards lossless CNNs with low-precision weights . In Proceedings of the International Conference on Learning Representations (ICLR\u201917) . Aojun Zhou, Anbang Yao, Yiwen Guo, Lin Xu, and Yurong Chen. 2017. Incremental network quantization: Towards lossless CNNs with low-precision weights. In Proceedings of the International Conference on Learning Representations (ICLR\u201917)."},{"key":"e_1_2_1_33_1","unstructured":"Shuchang Zhou Yuxin Wu Zekun Ni Xinyu Zhou He Wen and Yuheng Zou. 2016. DoReFa-Net: Training low bitwidth convolutional neural networks with low bitwidth gradients. arXiv:1606.06160  Shuchang Zhou Yuxin Wu Zekun Ni Xinyu Zhou He Wen and Yuheng Zou. 2016. DoReFa-Net: Training low bitwidth convolutional neural networks with low bitwidth gradients. arXiv:1606.06160"},{"key":"e_1_2_1_34_1","doi-asserted-by":"publisher","DOI":"10.1007\/s11390-017-1750-y"},{"key":"e_1_2_1_35_1","volume-title":"Proceedings of the International Conference on Learning Representations (ICLR\u201916)","author":"Zhu Chenzhuo","unstructured":"Chenzhuo Zhu , Song Han , Huizi Mao , and William J. Dally . 2016. Trained ternary quantization . In Proceedings of the International Conference on Learning Representations (ICLR\u201916) . Chenzhuo Zhu, Song Han, Huizi Mao, and William J. Dally. 2016. Trained ternary quantization. In Proceedings of the International Conference on Learning Representations (ICLR\u201916)."}],"container-title":["ACM Transactions on Computer Systems"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3444943","content-type":"unspecified","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/dl.acm.org\/doi\/pdf\/10.1145\/3444943","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,6,17]],"date-time":"2025-06-17T17:28:13Z","timestamp":1750181293000},"score":1,"resource":{"primary":{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3444943"}},"subtitle":["Uniform Noise Injection for Non-Uniform Quantization of Neural Networks"],"short-title":[],"issued":{"date-parts":[[2019,11,30]]},"references-count":35,"journal-issue":{"issue":"1-4","published-print":{"date-parts":[[2019,11,30]]}},"alternative-id":["10.1145\/3444943"],"URL":"https:\/\/doi.org\/10.1145\/3444943","relation":{},"ISSN":["0734-2071","1557-7333"],"issn-type":[{"value":"0734-2071","type":"print"},{"value":"1557-7333","type":"electronic"}],"subject":[],"published":{"date-parts":[[2019,11,30]]},"assertion":[{"value":"2020-06-01","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2020-12-01","order":1,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2021-03-26","order":2,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}