Task relevant autoencoding enhances machine learning for human neuroscience
Abstract In human neuroscience, machine learning can help reveal lower-dimensional neural representations relevant to subjects’ behavior. However, state-of-the-art models typically require large datasets to train, and so are prone to overfitting on human neuroimaging data that often possess few samp...
Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-01-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-024-83867-6 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841544661309587456 |
|---|---|
| author | Seyedmehdi Orouji Vincent Taschereau-Dumouchel Aurelio Cortese Brian Odegaard Cody Cushing Mouslim Cherkaoui Mitsuo Kawato Hakwan Lau Megan A. K. Peters |
| author_facet | Seyedmehdi Orouji Vincent Taschereau-Dumouchel Aurelio Cortese Brian Odegaard Cody Cushing Mouslim Cherkaoui Mitsuo Kawato Hakwan Lau Megan A. K. Peters |
| author_sort | Seyedmehdi Orouji |
| collection | DOAJ |
| description | Abstract In human neuroscience, machine learning can help reveal lower-dimensional neural representations relevant to subjects’ behavior. However, state-of-the-art models typically require large datasets to train, and so are prone to overfitting on human neuroimaging data that often possess few samples but many input dimensions. Here, we capitalized on the fact that the features we seek in human neuroscience are precisely those relevant to subjects’ behavior rather than noise or other irrelevant factors. We thus developed a Task-Relevant Autoencoder via Classifier Enhancement (TRACE) designed to identify behaviorally-relevant target neural patterns. We benchmarked TRACE against a standard autoencoder and other models for two severely truncated machine learning datasets (to match the data typically available in functional magnetic resonance imaging [fMRI] data for an individual subject), then evaluated all models on fMRI data from 59 subjects who observed animals and objects. TRACE outperformed alternative models nearly unilaterally, showing up to 12% increased classification accuracy and up to 56% improvement in discovering “cleaner”, task-relevant representations. These results showcase TRACE’s potential for a wide variety of data related to human behavior. |
| format | Article |
| id | doaj-art-51a3dae1e312465eaa2dd367dfee6bbd |
| institution | Kabale University |
| issn | 2045-2322 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Scientific Reports |
| spelling | doaj-art-51a3dae1e312465eaa2dd367dfee6bbd2025-01-12T12:21:04ZengNature PortfolioScientific Reports2045-23222025-01-0115111410.1038/s41598-024-83867-6Task relevant autoencoding enhances machine learning for human neuroscienceSeyedmehdi Orouji0Vincent Taschereau-Dumouchel1Aurelio Cortese2Brian Odegaard3Cody Cushing4Mouslim Cherkaoui5Mitsuo Kawato6Hakwan Lau7Megan A. K. Peters8Department of Cognitive Sciences, University of California, 2201 Social & Behavioral Sciences GatewayDepartment of Psychiatry and Addictology, Université de MontréalATR Computational Neuroscience LaboratoriesDepartment of Psychology, University of FloridaDepartment of Psychology, University of California Los AngelesDepartment of Psychology, University of California Los AngelesATR Computational Neuroscience LaboratoriesRIKEN Center for Brain ScienceDepartment of Cognitive Sciences, University of California, 2201 Social & Behavioral Sciences GatewayAbstract In human neuroscience, machine learning can help reveal lower-dimensional neural representations relevant to subjects’ behavior. However, state-of-the-art models typically require large datasets to train, and so are prone to overfitting on human neuroimaging data that often possess few samples but many input dimensions. Here, we capitalized on the fact that the features we seek in human neuroscience are precisely those relevant to subjects’ behavior rather than noise or other irrelevant factors. We thus developed a Task-Relevant Autoencoder via Classifier Enhancement (TRACE) designed to identify behaviorally-relevant target neural patterns. We benchmarked TRACE against a standard autoencoder and other models for two severely truncated machine learning datasets (to match the data typically available in functional magnetic resonance imaging [fMRI] data for an individual subject), then evaluated all models on fMRI data from 59 subjects who observed animals and objects. TRACE outperformed alternative models nearly unilaterally, showing up to 12% increased classification accuracy and up to 56% improvement in discovering “cleaner”, task-relevant representations. These results showcase TRACE’s potential for a wide variety of data related to human behavior.https://doi.org/10.1038/s41598-024-83867-6Human neuroscienceMachine learningDimensionality reductionTask-relevant representationfMRIMVPA |
| spellingShingle | Seyedmehdi Orouji Vincent Taschereau-Dumouchel Aurelio Cortese Brian Odegaard Cody Cushing Mouslim Cherkaoui Mitsuo Kawato Hakwan Lau Megan A. K. Peters Task relevant autoencoding enhances machine learning for human neuroscience Scientific Reports Human neuroscience Machine learning Dimensionality reduction Task-relevant representation fMRI MVPA |
| title | Task relevant autoencoding enhances machine learning for human neuroscience |
| title_full | Task relevant autoencoding enhances machine learning for human neuroscience |
| title_fullStr | Task relevant autoencoding enhances machine learning for human neuroscience |
| title_full_unstemmed | Task relevant autoencoding enhances machine learning for human neuroscience |
| title_short | Task relevant autoencoding enhances machine learning for human neuroscience |
| title_sort | task relevant autoencoding enhances machine learning for human neuroscience |
| topic | Human neuroscience Machine learning Dimensionality reduction Task-relevant representation fMRI MVPA |
| url | https://doi.org/10.1038/s41598-024-83867-6 |
| work_keys_str_mv | AT seyedmehdiorouji taskrelevantautoencodingenhancesmachinelearningforhumanneuroscience AT vincenttaschereaudumouchel taskrelevantautoencodingenhancesmachinelearningforhumanneuroscience AT aureliocortese taskrelevantautoencodingenhancesmachinelearningforhumanneuroscience AT brianodegaard taskrelevantautoencodingenhancesmachinelearningforhumanneuroscience AT codycushing taskrelevantautoencodingenhancesmachinelearningforhumanneuroscience AT mouslimcherkaoui taskrelevantautoencodingenhancesmachinelearningforhumanneuroscience AT mitsuokawato taskrelevantautoencodingenhancesmachinelearningforhumanneuroscience AT hakwanlau taskrelevantautoencodingenhancesmachinelearningforhumanneuroscience AT meganakpeters taskrelevantautoencodingenhancesmachinelearningforhumanneuroscience |