Feature Selection Methods for Zero-Shot Learning of Neural Activity
Dimensionality poses a serious challenge when making predictions from human neuroimaging data. Across imaging modalities, large pools of potential neural features (e.g., responses from particular voxels, electrodes, and temporal windows) have to be related to typically limited sets of stimuli and sa...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2017-06-01
|
| Series: | Frontiers in Neuroinformatics |
| Subjects: | |
| Online Access: | http://journal.frontiersin.org/article/10.3389/fninf.2017.00041/full |
| _version_ | 1811237477716852736 |
|---|---|
| author | Carlos A. Caceres Matthew J. Roos Kyle M. Rupp Griffin Milsap Nathan E. Crone Michael E. Wolmetz Christopher R. Ratto |
| author_facet | Carlos A. Caceres Matthew J. Roos Kyle M. Rupp Griffin Milsap Nathan E. Crone Michael E. Wolmetz Christopher R. Ratto |
| author_sort | Carlos A. Caceres |
| collection | DOAJ |
| description | Dimensionality poses a serious challenge when making predictions from human neuroimaging data. Across imaging modalities, large pools of potential neural features (e.g., responses from particular voxels, electrodes, and temporal windows) have to be related to typically limited sets of stimuli and samples. In recent years, zero-shot prediction models have been introduced for mapping between neural signals and semantic attributes, which allows for classification of stimulus classes not explicitly included in the training set. While choices about feature selection can have a substantial impact when closed-set accuracy, open-set robustness, and runtime are competing design objectives, no systematic study of feature selection for these models has been reported. Instead, a relatively straightforward feature stability approach has been adopted and successfully applied across models and imaging modalities. To characterize the tradeoffs in feature selection for zero-shot learning, we compared correlation-based stability to several other feature selection techniques on comparable data sets from two distinct imaging modalities: functional Magnetic Resonance Imaging and Electrocorticography. While most of the feature selection methods resulted in similar zero-shot prediction accuracies and spatial/spectral patterns of selected features, there was one exception; A novel feature/attribute correlation approach was able to achieve those accuracies with far fewer features, suggesting the potential for simpler prediction models that yield high zero-shot classification accuracy. |
| first_indexed | 2024-04-12T12:24:07Z |
| format | Article |
| id | doaj.art-44cd86fcba174b11bf0131bf68ca1356 |
| institution | Directory Open Access Journal |
| issn | 1662-5196 |
| language | English |
| last_indexed | 2024-04-12T12:24:07Z |
| publishDate | 2017-06-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Neuroinformatics |
| spelling | doaj.art-44cd86fcba174b11bf0131bf68ca13562022-12-22T03:33:12ZengFrontiers Media S.A.Frontiers in Neuroinformatics1662-51962017-06-011110.3389/fninf.2017.00041236545Feature Selection Methods for Zero-Shot Learning of Neural ActivityCarlos A. Caceres0Matthew J. Roos1Kyle M. Rupp2Griffin Milsap3Nathan E. Crone4Michael E. Wolmetz5Christopher R. Ratto6Applied Physics Laboratory, Johns Hopkins UniversityLaurel, MD, United StatesApplied Physics Laboratory, Johns Hopkins UniversityLaurel, MD, United StatesDepartment of Biomedical Engineering, Johns Hopkins UniversityBaltimore, MD, United StatesDepartment of Biomedical Engineering, Johns Hopkins UniversityBaltimore, MD, United StatesDepartment of Neurology, Johns Hopkins MedicineBaltimore, MD, United StatesApplied Physics Laboratory, Johns Hopkins UniversityLaurel, MD, United StatesApplied Physics Laboratory, Johns Hopkins UniversityLaurel, MD, United StatesDimensionality poses a serious challenge when making predictions from human neuroimaging data. Across imaging modalities, large pools of potential neural features (e.g., responses from particular voxels, electrodes, and temporal windows) have to be related to typically limited sets of stimuli and samples. In recent years, zero-shot prediction models have been introduced for mapping between neural signals and semantic attributes, which allows for classification of stimulus classes not explicitly included in the training set. While choices about feature selection can have a substantial impact when closed-set accuracy, open-set robustness, and runtime are competing design objectives, no systematic study of feature selection for these models has been reported. Instead, a relatively straightforward feature stability approach has been adopted and successfully applied across models and imaging modalities. To characterize the tradeoffs in feature selection for zero-shot learning, we compared correlation-based stability to several other feature selection techniques on comparable data sets from two distinct imaging modalities: functional Magnetic Resonance Imaging and Electrocorticography. While most of the feature selection methods resulted in similar zero-shot prediction accuracies and spatial/spectral patterns of selected features, there was one exception; A novel feature/attribute correlation approach was able to achieve those accuracies with far fewer features, suggesting the potential for simpler prediction models that yield high zero-shot classification accuracy.http://journal.frontiersin.org/article/10.3389/fninf.2017.00041/fullzero-shot learningtransfer learningsemanticsfMRIelectrocorticographyfeature selection |
| spellingShingle | Carlos A. Caceres Matthew J. Roos Kyle M. Rupp Griffin Milsap Nathan E. Crone Michael E. Wolmetz Christopher R. Ratto Feature Selection Methods for Zero-Shot Learning of Neural Activity Frontiers in Neuroinformatics zero-shot learning transfer learning semantics fMRI electrocorticography feature selection |
| title | Feature Selection Methods for Zero-Shot Learning of Neural Activity |
| title_full | Feature Selection Methods for Zero-Shot Learning of Neural Activity |
| title_fullStr | Feature Selection Methods for Zero-Shot Learning of Neural Activity |
| title_full_unstemmed | Feature Selection Methods for Zero-Shot Learning of Neural Activity |
| title_short | Feature Selection Methods for Zero-Shot Learning of Neural Activity |
| title_sort | feature selection methods for zero shot learning of neural activity |
| topic | zero-shot learning transfer learning semantics fMRI electrocorticography feature selection |
| url | http://journal.frontiersin.org/article/10.3389/fninf.2017.00041/full |
| work_keys_str_mv | AT carlosacaceres featureselectionmethodsforzeroshotlearningofneuralactivity AT matthewjroos featureselectionmethodsforzeroshotlearningofneuralactivity AT kylemrupp featureselectionmethodsforzeroshotlearningofneuralactivity AT griffinmilsap featureselectionmethodsforzeroshotlearningofneuralactivity AT nathanecrone featureselectionmethodsforzeroshotlearningofneuralactivity AT michaelewolmetz featureselectionmethodsforzeroshotlearningofneuralactivity AT christopherrratto featureselectionmethodsforzeroshotlearningofneuralactivity |