{"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/taco-learning-task-decomposition-via-temporal","title":"TACO: Learning Task Decomposition via Temporal Alignment for Control","arxiv_id":"1803.01840","date":"2018-03-02","proceeding":"ICML 2018 7","authors":["Kyriacos Shiarlis","Markus Wulfmeier","Sasha Salter","Shimon Whiteson","Ingmar Posner"],"abstract":"Many advanced Learning from Demonstration (LfD) methods consider the\ndecomposition of complex, real-world tasks into simpler sub-tasks. By reusing\nthe corresponding sub-policies within and between tasks, they provide training\ndata for each policy from different high-level tasks and compose them to\nperform novel ones. Existing approaches to modular LfD focus either on learning\na single high-level task or depend on domain knowledge and temporal\nsegmentation. In contrast, we propose a weakly supervised, domain-agnostic\napproach based on task sketches, which include only the sequence of sub-tasks\nperformed in each demonstration. Our approach simultaneously aligns the\nsketches with the observed demonstrations and learns the required sub-policies.\nThis improves generalisation in comparison to separate optimisation procedures.\nWe evaluate the approach on multiple domains, including a simulated 3D robot\narm control task using purely image-based observations. The results show that\nour approach performs commensurately with fully supervised approaches, while\nrequiring significantly less annotation effort.","url_abs":"http://arxiv.org/abs/1803.01840v2","url_pdf":"http://arxiv.org/pdf/1803.01840v2.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":"taco-learning-task-decomposition-via-temporal","repo_url":"https://github.com/KyriacosShiarli/taco","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"GPL-3.0"}}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1803.01840","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}