Research map / Label-efficient and robust learning / Semi-supervised learning
Constructing Multiple Tasks for Augmentation: Improving Neural Image Classification with K-Means Features
Tao Gui, Lizhi Qing, Qi Zhang, Jiacheng Ye, Hang Yan, Zichu Fei, Xuanjing Huang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add Meta-MTL with k-means auxiliary tasks | STL classifier · Omniglot | accuracy: 65.72 -> 74.80 | Improved |
| Add Meta-MTL auxiliary-task training | STL classifier · MNIST, 1% training data | accuracy: 91.83 -> 94.22 | Improved |
| Use the remaining unlabeled data to train embeddings (ablation) | Meta-MTL, MNIST · MNIST, 1% training data | accuracy: 94.22 -> 94.42 | Improved |
| Add Meta-MTL with k-means auxiliary tasks | STL classifier · CIFAR-10 | accuracy: 72.48 -> 75.69 | Improved |
| Add Meta-MTL with k-means auxiliary tasks | STL classifier with data augmentation · CIFAR-10, CNN with data augmentation | accuracy: 75.91 -> 79.65 | Improved |
| Add Meta-MTL with k-means auxiliary tasks | STL classifier · CIFAR-100, 20 C | accuracy: 55.94 -> 60.39 | Improved |
| Add Meta-MTL with k-means auxiliary tasks | STL classifier · CIFAR-100, 100 C | accuracy: 44.19 -> 47.94 | Improved |
| Add Meta-MTL to k-means task training (ablation) | MTL, k-means labels, miniImageNet DeepCluster embeddings · miniImageNet | accuracy: 37.24 -> 41.07 | Improved |
| Use DeepCluster rather than ACAI embeddings (ablation) | Meta-MTL, random scaling · miniImageNet | accuracy: 40.43 -> 41.07 | Improved |
| Increase auxiliary tasks from T = 4 to T = 12 (ablation) | Meta-MTL · CIFAR-10 | accuracy: 75.02 -> 75.69 | Improved |
| Use k-means labels rather than random labels (ablation) | Meta-MTL · CIFAR-10, varying number of auxiliary tasks | accuracy: more than 4% | Improved |
| Replace Meta-MTL with ordinary MTL on k-means tasks (ablation) | CIFAR-10 · CIFAR-10, varying number of auxiliary tasks | accuracy: more than 8% | Hurt |
Closest papers by meaning
- Unsupervised Learning via Meta-Learning (ICLR 2019 · similarity 0.88)
- Clustered Task-Aware Meta-Learning by Learning From Learning Paths (TPAMI 2023 · similarity 0.88)
- Bridging Multi-Task Learning and Meta-Learning: Towards Efficient Training and Effective Adaptation (ICML 2021 · similarity 0.88)
- Towards Principled Task Grouping for Multi-Task Learning (NeurIPS 2026 · similarity 0.86)
- Joint-Task Regularization for Partially Labeled Multi-Task Learning (CVPR 2024 · similarity 0.85)
- Distribution Matching for Multi-Task Learning of Classification Tasks: A Large-Scale Study on Faces & Beyond (AAAI 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
