{"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/visual-pathways-from-the-perspective-of-cost","title":"Visual pathways from the perspective of cost functions and multi-task deep neural networks","arxiv_id":"1706.01757","date":"2017-09-16","proceeding":null,"authors":[],"abstract":"Vision research has been shaped by the seminal insight that we can understand\nthe higher-tier visual cortex from the perspective of multiple functional\npathways with different goals. In this paper, we try to give a computational\naccount of the functional organization of this system by reasoning from the\nperspective of multi-task deep neural networks. Machine learning has shown that\ntasks become easier to solve when they are decomposed into subtasks with their\nown cost function. We hypothesize that the visual system optimizes multiple\ncost functions of unrelated tasks and this causes the emergence of a ventral\npathway dedicated to vision for perception, and a dorsal pathway dedicated to\nvision for action. To evaluate the functional organization in multi-task deep\nneural networks, we propose a method that measures the contribution of a unit\ntowards each task, applying it to two networks that have been trained on either\ntwo related or two unrelated tasks, using an identical stimulus set. Results\nshow that the network trained on the unrelated tasks shows a decreasing degree\nof feature representation sharing towards higher-tier layers while the network\ntrained on related tasks uniformly shows high degree of sharing. We conjecture\nthat the method we propose can be used to analyze the anatomical and functional\norganization of the visual system and beyond. We predict that the degree to\nwhich tasks are related is a good descriptor of the degree to which they share\ndownstream cortical-units.","url_abs":"http://arxiv.org/abs/1706.01757v2","url_pdf":"http://arxiv.org/pdf/1706.01757v2.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":"visual-pathways-from-the-perspective-of-cost","repo_url":"https://github.com/mlosch/FeatureSharing","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}