{"id":17302116,"url":"https://github.com/tanyuqian/learning-data-manipulation","last_synced_at":"2025-10-26T16:13:11.388Z","repository":{"id":134189534,"uuid":"217189507","full_name":"tanyuqian/learning-data-manipulation","owner":"tanyuqian","description":"NeurIPS 2019 - Learning Data Manipulation for Augmentation and Weighting","archived":false,"fork":false,"pushed_at":"2020-09-05T23:17:25.000Z","size":77,"stargazers_count":109,"open_issues_count":5,"forks_count":16,"subscribers_count":4,"default_branch":"master","last_synced_at":"2025-04-14T12:57:15.473Z","etag":null,"topics":["bert","data-augmentation","data-manipulation","meta-learning"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/tanyuqian.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2019-10-24T01:56:10.000Z","updated_at":"2024-10-21T08:18:03.000Z","dependencies_parsed_at":null,"dependency_job_id":"45990c5c-86ad-41de-b610-37619728e635","html_url":"https://github.com/tanyuqian/learning-data-manipulation","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/tanyuqian%2Flearning-data-manipulation","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/tanyuqian%2Flearning-data-manipulation/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/tanyuqian%2Flearning-data-manipulation/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/tanyuqian%2Flearning-data-manipulation/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/tanyuqian","download_url":"https://codeload.github.com/tanyuqian/learning-data-manipulation/tar.gz/refs/heads/master","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/tanyuqian%2Flearning-data-manipulation/sbom","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":259172111,"owners_count":22816514,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["bert","data-augmentation","data-manipulation","meta-learning"],"created_at":"2024-10-15T11:46:37.518Z","updated_at":"2025-10-26T16:13:06.338Z","avatar_url":"https://github.com/tanyuqian.png","language":"Python","readme":"# Learning Data Manipulation\n\nThis repo contains preliminary code of the following paper:\n\n[Learning Data Manipulation for Augmentation and Weighting](http://www.cs.cmu.edu/~zhitingh/data/neurips19_data_manip_preprint.pdf)  \nZhiting Hu*, Bowen Tan*, Ruslan Salakhutdinov, Tom Mitchell, Eric P. Xing  \nNeurIPS 2019 (equal contribution)\n\n## Requirements\n\n- `python3.6`\n- `pytorch==1.0.1`\n- `pytorch_pretrained_bert==0.6.1`\n- `torchvision==0.2.2`\n\n## Code\n* ```baseline_main.py```: Vanilla BERT Classifier.\n* ```ren_main.py```: Described in [(Ren et al.)](https://arxiv.org/pdf/1803.09050.pdf).\n* ```weighting_main.py```: Our weighting algorithm.\n* ```augmentation_main.py```: Our augmentation algorithm.\n\n\n## Running\nRunning scripts for experiments are available in [scripts/](scripts/).\n\n## Results\n\nAll the detailed training logs are availble in [results/](results/).\n\n*(Note: The result numbers may be slightly different from those in the paper due to slightly different implementation details and random seeds, while the improvements over comparison methods are consistent.)*\n\n### low data\n\n##### SST-5\n|Base Model: BERT|Ren et al.| Weighting  | Augmentation |\n|:-:|:-:|:-:|:-:|\n| 33.32 ± 4.04 | 36.09 ± 2.26 | 36.51 ± 2.54   | 37.55 ± 2.63 |\n\n##### CIFAR-10\n|                  |  Pretrained    | Not Pretrained |\n|------------------|----------------|----------------|\n|Base Model: ResNet| 34.58 ± 4.13   | 24.68 ± 3.29   |\n| Ren et al.       | 23.29 ± 5.95   | 22.26 ± 2.80   |\n| Weighting        | 36.75 ± 3.09   | 26.47 ± 1.69   |\n\n\n### imbalanced data\n\n##### SST-2\n|| 20 : 1000 | 50 : 1000　| 100 : 1000\n|:-:|:-:|:-:|:-:|\n|Base Model: BERT| 54.91 ± 5.98 | 67.73 ± 9.20 | 75.04 ± 4.51 |\n|Ren et al.| 74.61 ± 3.54 | 76.89 ± 5.07 | 80.73 ± 2.19 | \n|Weighting| 75.08 ± 4.98 | 79.35 ± 2.59 | 81.82 ± 1.88 | \n\n##### CIFAR-10\n|                  | 20 : 1000    | 50 : 1000    | 100 : 1000   |\n|------------------|--------------|--------------|--------------|\n|Base Model: ResNet| 70.65 ± 4.98 | 79.52 ± 4.81 | 86.12 ± 3.37 |\n| Ren et al.       | 76.68 ± 5.35 | 77.34 ± 7.38 | 78.57 ± 5.61 |\n| Weighting        | 79.07 ± 5.02 | 82.65 ± 5.13 | 87.63 ± 3.72 |\n","funding_links":[],"categories":[],"sub_categories":[],"project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ftanyuqian%2Flearning-data-manipulation","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Ftanyuqian%2Flearning-data-manipulation","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ftanyuqian%2Flearning-data-manipulation/lists"}