International Conference on Machine Learning (ICML), 2024

[Paper] | [Poster] | [Uncertainty Analysis] | [Citation]

Abstract: Uncertainty estimation (UE), as an effective means of quantifying predictive uncertainty, is crucial for safe and reliable decision-making, especially in high-risk scenarios. Existing UE schemes usually assume that there are completely-labeled samples to support fully-supervised learning. In practice, however, many UE tasks often have no sufficiently-labeled data to use, such as the Multiple Instance Learning (MIL) with only weak instance annotations. To bridge this gap, this paper, for the first time, addresses the weakly-supervised issue of Multi-Instance UE (MIUE) and proposes a new baseline scheme, Multi-Instance Residual Evidential Learning (MIREL). Particularly, at the fine-grained instance UE with only weak supervision, we derive a multi-instance residual operator through the Fundamental Theorem of Symmetric Functions. On this operator derivation, we further propose MIREL to jointly model the high-order predictive distribution at bag and instance levels for MIUE. Extensive experiments empirically demonstrate that our MIREL not only could often make existing MIL networks perform better in MIUE, but also could surpass representative UE methods by large margins, especially in instance-level UE tasks.

📚 Recent updates:

• 24/06/30: Updated the link to paper and uploaded poster.
• 24/05/28: Uploaded experimental files, Notebook - Bag-level-Uncertainty-Analysis, and Notebook - Instance-level-Uncertainty-Analysis.
• 24/05/27: Uploaded MIREL source codes.

Use the following command to load configurations from a YAML file and train the model:

python3 main.py --config config/mnist-bags/cfg_abmil_mirel.yml --handler clf --multi_run

The configuration files for running MIREL models with different bag datasets (MNIST-Bags, CIFAR10-Bags, and CAMELYON16) are provided here. A detailed description of each configuration is commented on in these files.

All experimental files can be found in here.

Please check

• Bag-level uncertainty analysis: Notebook - Bag-level-Uncertainty-Analysis
• Instance-level uncertainty analysis: Notebook - Instance-level-Uncertainty-Analysis

If you find this work helps you more or less, please cite it via

@InProceedings{pmlr-v235-liu24ac,
    title = {Weakly-Supervised Residual Evidential Learning for Multi-Instance Uncertainty Estimation},
    author = {Liu, Pei and Ji, Luping},
    booktitle = {Proceedings of the 41st International Conference on Machine Learning},
    pages = {31262--31292},
    year = {2024},
    volume = {235},
    publisher = {Proceedings of Machine Learning Research},
    url = {https://proceedings.mlr.press/v235/liu24ac.html}
}

Any issues can be sent via E-mail ([email protected]) or posted on the issue page.