{"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/deep-learning-as-a-tool-for-neural-data","title":"Deep learning as a tool for neural data analysis: speech classification and cross-frequency coupling in human sensorimotor cortex","arxiv_id":"1803.09807","date":"2018-03-26","proceeding":null,"authors":["Jesse A. Livezey","Kristofer E. Bouchard","Edward F. Chang"],"abstract":"A fundamental challenge in neuroscience is to understand what structure in\nthe world is represented in spatially distributed patterns of neural activity\nfrom multiple single-trial measurements. This is often accomplished by learning\na simple, linear transformations between neural features and features of the\nsensory stimuli or motor task. While successful in some early sensory\nprocessing areas, linear mappings are unlikely to be ideal tools for\nelucidating nonlinear, hierarchical representations of higher-order brain areas\nduring complex tasks, such as the production of speech by humans. Here, we\napply deep networks to predict produced speech syllables from cortical surface\nelectric potentials recorded from human sensorimotor cortex. We found that deep\nnetworks had higher decoding prediction accuracy compared to baseline models,\nand also exhibited greater improvements in accuracy with increasing dataset\nsize. We further demonstrate that deep network's confusions revealed\nhierarchical latent structure in the neural data, which recapitulated the\nunderlying articulatory nature of speech motor control. Finally, we used deep\nnetworks to compare task-relevant information in different neural frequency\nbands, and found that the high-gamma band contains the vast majority of\ninformation relevant for the speech prediction task, with little-to-no\nadditional contribution from lower-frequencies. Together, these results\ndemonstrate the utility of deep networks as a data analysis tool for\nneuroscience.","url_abs":"http://arxiv.org/abs/1803.09807v1","url_pdf":"http://arxiv.org/pdf/1803.09807v1.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":"deep-learning-as-a-tool-for-neural-data","repo_url":"https://github.com/BouchardLab/deprecated_process_ecog","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"deep-learning-as-a-tool-for-neural-data","repo_url":"https://github.com/BouchardLab/process_ecog","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"classification","task_name":"General Classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}