Noise contrastive meta-learning for conditional density estimation using kernel mean embeddings
Current meta-learning approaches focus on learning functional representations of relationships between variables, i.e. estimating conditional expectations in regression. In many applications, however, the conditional distributions cannot be meaningfully summarized solely by expectation (due to e.g....
| Main Authors: | , , , |
|---|---|
| Format: | Journal article |
| Language: | English |
| Published: |
Journal of Machine Learning Research
2021
|
| _version_ | 1797080674036875264 |
|---|---|
| author | Ton, J-F Chan, L Teh, YW Sejdinovic, D |
| author_facet | Ton, J-F Chan, L Teh, YW Sejdinovic, D |
| author_sort | Ton, J-F |
| collection | OXFORD |
| description | Current meta-learning approaches focus on learning functional representations of relationships between variables, i.e. estimating conditional expectations in regression. In many applications, however, the conditional distributions cannot be meaningfully summarized solely by expectation (due to e.g. multimodality). We introduce a novel technique for meta-learning conditional densities, which combines neural representation and noise contrastive estimation together with well-established literature in conditional mean embeddings into reproducing kernel Hilbert spaces. The method shows significant improvements over standard density estimation methods on synthetic and real-world data, by leveraging shared representations across multiple conditional density estimation tasks. |
| first_indexed | 2024-03-07T01:03:28Z |
| format | Journal article |
| id | oxford-uuid:8a82cd96-2c5f-4d64-9562-51240cc95a71 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T01:03:28Z |
| publishDate | 2021 |
| publisher | Journal of Machine Learning Research |
| record_format | dspace |
| spelling | oxford-uuid:8a82cd96-2c5f-4d64-9562-51240cc95a712022-03-26T22:32:01ZNoise contrastive meta-learning for conditional density estimation using kernel mean embeddingsJournal articlehttp://purl.org/coar/resource_type/c_dcae04bcuuid:8a82cd96-2c5f-4d64-9562-51240cc95a71EnglishSymplectic ElementsJournal of Machine Learning Research2021Ton, J-FChan, LTeh, YWSejdinovic, DCurrent meta-learning approaches focus on learning functional representations of relationships between variables, i.e. estimating conditional expectations in regression. In many applications, however, the conditional distributions cannot be meaningfully summarized solely by expectation (due to e.g. multimodality). We introduce a novel technique for meta-learning conditional densities, which combines neural representation and noise contrastive estimation together with well-established literature in conditional mean embeddings into reproducing kernel Hilbert spaces. The method shows significant improvements over standard density estimation methods on synthetic and real-world data, by leveraging shared representations across multiple conditional density estimation tasks. |
| spellingShingle | Ton, J-F Chan, L Teh, YW Sejdinovic, D Noise contrastive meta-learning for conditional density estimation using kernel mean embeddings |
| title | Noise contrastive meta-learning for conditional density estimation using kernel mean embeddings |
| title_full | Noise contrastive meta-learning for conditional density estimation using kernel mean embeddings |
| title_fullStr | Noise contrastive meta-learning for conditional density estimation using kernel mean embeddings |
| title_full_unstemmed | Noise contrastive meta-learning for conditional density estimation using kernel mean embeddings |
| title_short | Noise contrastive meta-learning for conditional density estimation using kernel mean embeddings |
| title_sort | noise contrastive meta learning for conditional density estimation using kernel mean embeddings |
| work_keys_str_mv | AT tonjf noisecontrastivemetalearningforconditionaldensityestimationusingkernelmeanembeddings AT chanl noisecontrastivemetalearningforconditionaldensityestimationusingkernelmeanembeddings AT tehyw noisecontrastivemetalearningforconditionaldensityestimationusingkernelmeanembeddings AT sejdinovicd noisecontrastivemetalearningforconditionaldensityestimationusingkernelmeanembeddings |