{"scopus_import":"1","year":"2021","author":[{"full_name":"Baskin, Chaim","last_name":"Baskin","first_name":"Chaim"},{"first_name":"Natan","last_name":"Liss","full_name":"Liss, Natan"},{"full_name":"Schwartz, Eli","first_name":"Eli","last_name":"Schwartz"},{"full_name":"Zheltonozhskii, Evgenii","first_name":"Evgenii","last_name":"Zheltonozhskii"},{"first_name":"Raja","last_name":"Giryes","full_name":"Giryes, Raja"},{"first_name":"Alexander","orcid":"0000-0001-9699-8730","last_name":"Bronstein","full_name":"Bronstein, Alexander","id":"58f3726e-7cba-11ef-ad8b-e6e8cb3904e6"},{"full_name":"Mendelson, Avi","last_name":"Mendelson","first_name":"Avi"}],"language":[{"iso":"eng"}],"publication_status":"published","citation":{"mla":"Baskin, Chaim, et al. “UNIQ: Uniform Noise Injection for Non-Uniform Quantization of Neural Networks.” <i>ACM Transactions on Computer Systems</i>, vol. 37, no. 1–4, Association for Computing Machinery, 2021, pp. 1–15, doi:<a href=\"https://doi.org/10.1145/3444943\">10.1145/3444943</a>.","apa":"Baskin, C., Liss, N., Schwartz, E., Zheltonozhskii, E., Giryes, R., Bronstein, A. M., &#38; Mendelson, A. (2021). UNIQ: Uniform Noise Injection for Non-Uniform Quantization of neural networks. <i>ACM Transactions on Computer Systems</i>. Association for Computing Machinery. <a href=\"https://doi.org/10.1145/3444943\">https://doi.org/10.1145/3444943</a>","ieee":"C. Baskin <i>et al.</i>, “UNIQ: Uniform Noise Injection for Non-Uniform Quantization of neural networks,” <i>ACM Transactions on Computer Systems</i>, vol. 37, no. 1–4. Association for Computing Machinery, pp. 1–15, 2021.","ista":"Baskin C, Liss N, Schwartz E, Zheltonozhskii E, Giryes R, Bronstein AM, Mendelson A. 2021. UNIQ: Uniform Noise Injection for Non-Uniform Quantization of neural networks. ACM Transactions on Computer Systems. 37(1–4), 1–15.","chicago":"Baskin, Chaim, Natan Liss, Eli Schwartz, Evgenii Zheltonozhskii, Raja Giryes, Alex M. Bronstein, and Avi Mendelson. “UNIQ: Uniform Noise Injection for Non-Uniform Quantization of Neural Networks.” <i>ACM Transactions on Computer Systems</i>. Association for Computing Machinery, 2021. <a href=\"https://doi.org/10.1145/3444943\">https://doi.org/10.1145/3444943</a>.","short":"C. Baskin, N. Liss, E. Schwartz, E. Zheltonozhskii, R. Giryes, A.M. Bronstein, A. Mendelson, ACM Transactions on Computer Systems 37 (2021) 1–15.","ama":"Baskin C, Liss N, Schwartz E, et al. UNIQ: Uniform Noise Injection for Non-Uniform Quantization of neural networks. <i>ACM Transactions on Computer Systems</i>. 2021;37(1-4):1-15. doi:<a href=\"https://doi.org/10.1145/3444943\">10.1145/3444943</a>"},"date_published":"2021-03-26T00:00:00Z","status":"public","article_type":"original","month":"03","_id":"18237","article_processing_charge":"No","date_updated":"2024-10-15T07:47:22Z","issue":"1-4","title":"UNIQ: Uniform Noise Injection for Non-Uniform Quantization of neural networks","external_id":{"arxiv":["1804.10969"]},"abstract":[{"lang":"eng","text":"We present a novel method for neural network quantization. Our method, named UNIQ, emulates a non-uniform k-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."}],"extern":"1","date_created":"2024-10-08T12:58:26Z","intvolume":"        37","type":"journal_article","page":"1-15","oa":1,"OA_type":"green","volume":37,"publication_identifier":{"issn":["0734-2071"],"eissn":["1557-7333"]},"quality_controlled":"1","main_file_link":[{"open_access":"1","url":"https://doi.org/10.48550/arXiv.1804.10969"}],"publisher":"Association for Computing Machinery","publication":"ACM Transactions on Computer Systems","day":"26","oa_version":"Preprint","OA_place":"repository","doi":"10.1145/3444943","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87"}