{"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/input-fast-forwarding-for-better-deep","title":"Input Fast-Forwarding for Better Deep Learning","arxiv_id":"1705.08479","date":"2017-05-23","proceeding":null,"authors":["Ahmed Ibrahim","A. Lynn Abbott","Mohamed E. Hussein"],"abstract":"This paper introduces a new architectural framework, known as input\nfast-forwarding, that can enhance the performance of deep networks. The main\nidea is to incorporate a parallel path that sends representations of input\nvalues forward to deeper network layers. This scheme is substantially different\nfrom \"deep supervision\" in which the loss layer is re-introduced to earlier\nlayers. The parallel path provided by fast-forwarding enhances the training\nprocess in two ways. First, it enables the individual layers to combine\nhigher-level information (from the standard processing path) with lower-level\ninformation (from the fast-forward path). Second, this new architecture reduces\nthe problem of vanishing gradients substantially because the fast-forwarding\npath provides a shorter route for gradient backpropagation. In order to\nevaluate the utility of the proposed technique, a Fast-Forward Network (FFNet),\nwith 20 convolutional layers along with parallel fast-forward paths, has been\ncreated and tested. The paper presents empirical results that demonstrate\nimproved learning capacity of FFNet due to fast-forwarding, as compared to\nGoogLeNet (with deep supervision) and CaffeNet, which are 4x and 18x larger in\nsize, respectively. All of the source code and deep learning models described\nin this paper will be made available to the entire research community","url_abs":"http://arxiv.org/abs/1705.08479v1","url_pdf":"http://arxiv.org/pdf/1705.08479v1.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":"input-fast-forwarding-for-better-deep","repo_url":"https://github.com/aicentral/FFNet","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"deep-learning","task_name":"Deep Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}