{"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/learning-a-wavelet-like-auto-encoder-to","title":"Learning a Wavelet-like Auto-Encoder to Accelerate Deep Neural Networks","arxiv_id":"1712.07493","date":"2017-12-20","proceeding":null,"authors":["Tianshui Chen","Liang Lin","WangMeng Zuo","Xiaonan Luo","Lei Zhang"],"abstract":"Accelerating deep neural networks (DNNs) has been attracting increasing\nattention as it can benefit a wide range of applications, e.g., enabling mobile\nsystems with limited computing resources to own powerful visual recognition\nability. A practical strategy to this goal usually relies on a two-stage\nprocess: operating on the trained DNNs (e.g., approximating the convolutional\nfilters with tensor decomposition) and fine-tuning the amended network, leading\nto difficulty in balancing the trade-off between acceleration and maintaining\nrecognition performance. In this work, aiming at a general and comprehensive\nway for neural network acceleration, we develop a Wavelet-like Auto-Encoder\n(WAE) that decomposes the original input image into two low-resolution channels\n(sub-images) and incorporate the WAE into the classification neural networks\nfor joint training. The two decomposed channels, in particular, are encoded to\ncarry the low-frequency information (e.g., image profiles) and high-frequency\n(e.g., image details or noises), respectively, and enable reconstructing the\noriginal input image through the decoding process. Then, we feed the\nlow-frequency channel into a standard classification network such as VGG or\nResNet and employ a very lightweight network to fuse with the high-frequency\nchannel to obtain the classification result. Compared to existing DNN\nacceleration solutions, our framework has the following advantages: i) it is\ntolerant to any existing convolutional neural networks for classification\nwithout amending their structures; ii) the WAE provides an interpretable way to\npreserve the main components of the input image for classification.","url_abs":"http://arxiv.org/abs/1712.07493v1","url_pdf":"http://arxiv.org/pdf/1712.07493v1.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":"learning-a-wavelet-like-auto-encoder-to","repo_url":"https://github.com/tianshuichen/Wavelet-like-Auto-Encoder","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":{"status":"unanswered"}},{"paper_slug":"learning-a-wavelet-like-auto-encoder-to","repo_url":"https://github.com/kartikeya-badola/Wavelet-like-autoencoder-PyTorch-","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"classification-1","task_name":"Classification"},{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"tensor-decomposition","task_name":"Tensor Decomposition"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1712.07493","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}