{"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/mt-cgcnn-integrating-crystal-graph","title":"MT-CGCNN: Integrating Crystal Graph Convolutional Neural Network with Multitask Learning for Material Property Prediction","arxiv_id":"1811.05660","date":"2018-11-14","proceeding":null,"authors":["Soumya Sanyal","Janakiraman Balachandran","Naganand Yadati","Abhishek Kumar","Padmini Rajagopalan","Suchismita Sanyal","Partha Talukdar"],"abstract":"Developing accurate, transferable and computationally inexpensive machine\nlearning models can rapidly accelerate the discovery and development of new\nmaterials. Some of the major challenges involved in developing such models are,\n(i) limited availability of materials data as compared to other fields, (ii)\nlack of universal descriptor of materials to predict its various properties.\nThe limited availability of materials data can be addressed through transfer\nlearning, while the generic representation was recently addressed by Xie and\nGrossman [1], where they developed a crystal graph convolutional neural network\n(CGCNN) that provides a unified representation of crystals. In this work, we\ndevelop a new model (MT-CGCNN) by integrating CGCNN with transfer learning\nbased on multi-task (MT) learning. We demonstrate the effectiveness of MT-CGCNN\nby simultaneous prediction of various material properties such as Formation\nEnergy, Band Gap and Fermi Energy for a wide range of inorganic crystals (46774\nmaterials). MT-CGCNN is able to reduce the test error when employed on\ncorrelated properties by upto 8%. The model prediction has lower test error\ncompared to CGCNN, even when the training data is reduced by 10%. We also\ndemonstrate our model's better performance through prediction of end user\nscenario related to metal/non-metal classification. These results encourage\nfurther development of machine learning approaches which leverage multi-task\nlearning to address the aforementioned challenges in the discovery of new\nmaterials. We make MT-CGCNN's source code available to encourage reproducible\nresearch.","url_abs":"http://arxiv.org/abs/1811.05660v1","url_pdf":"http://arxiv.org/pdf/1811.05660v1.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":"mt-cgcnn-integrating-crystal-graph","repo_url":"https://github.com/soumyasanyal/mt-cgcnn","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}}],"tasks":[{"task_slug":"machine-learning","task_name":"BIG-bench Machine Learning"},{"task_slug":"band-gap","task_name":"Band Gap"},{"task_slug":"formation-energy","task_name":"Formation Energy"},{"task_slug":"multi-task-learning","task_name":"Multi-Task Learning"},{"task_slug":"property-prediction","task_name":"Property Prediction"},{"task_slug":"transfer-learning","task_name":"Transfer Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/formation-energy-on-materials-project","task":"Formation Energy","dataset":"Materials Project","model":"MT-CGCNN","rank_in_archive_order":9,"of":9,"metrics":{"MAE":"41"},"uses_additional_data":false}],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1811.05660","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}