{"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/training-neural-networks-by-using-power","title":"Training Neural Networks by Using Power Linear Units (PoLUs)","arxiv_id":"1802.00212","date":"2018-02-01","proceeding":null,"authors":["Yikang Li","Pak Lun Kevin Ding","Baoxin Li"],"abstract":"In this paper, we introduce \"Power Linear Unit\" (PoLU) which increases the\nnonlinearity capacity of a neural network and thus helps improving its\nperformance. PoLU adopts several advantages of previously proposed activation\nfunctions. First, the output of PoLU for positive inputs is designed to be\nidentity to avoid the gradient vanishing problem. Second, PoLU has a non-zero\noutput for negative inputs such that the output mean of the units is close to\nzero, hence reducing the bias shift effect. Thirdly, there is a saturation on\nthe negative part of PoLU, which makes it more noise-robust for negative\ninputs. Furthermore, we prove that PoLU is able to map more portions of every\nlayer's input to the same space by using the power function and thus increases\nthe number of response regions of the neural network. We use image\nclassification for comparing our proposed activation function with others. In\nthe experiments, MNIST, CIFAR-10, CIFAR-100, Street View House Numbers (SVHN)\nand ImageNet are used as benchmark datasets. The neural networks we implemented\ninclude widely-used ELU-Network, ResNet-50, and VGG16, plus a couple of shallow\nnetworks. Experimental results show that our proposed activation function\noutperforms other state-of-the-art models with most networks.","url_abs":"http://arxiv.org/abs/1802.00212v1","url_pdf":"http://arxiv.org/pdf/1802.00212v1.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":"training-neural-networks-by-using-power","repo_url":"https://github.com/awur978/Autoencoder","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}}],"tasks":[{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}