On Characterizing and Mitigating Imbalances in Multi-Instance Partial Label Learning

• URL: http://arxiv.org/abs/2407.10000v1
• Date: Sat, 13 Jul 2024 20:56:34 GMT
• Title: On Characterizing and Mitigating Imbalances in Multi-Instance Partial Label Learning
• Authors: Kaifu Wang, Efthymia Tsamoura, Dan Roth,
• Abstract summary: Multi-Instance Partial Label Learning (MI-PLL) is a weakly-supervised learning setting. We focus on characterizing and mitigating learning imbalances under MI-PLL.
• Score: 57.18649648182171
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Multi-Instance Partial Label Learning (MI-PLL) is a weakly-supervised learning setting encompassing partial label learning, latent structural learning, and neurosymbolic learning. Differently from supervised learning, in MI-PLL, the inputs to the classifiers at training-time are tuples of instances $\textbf{x}$, while the supervision signal is generated by a function $\sigma$ over the gold labels of $\textbf{x}$. The gold labels are hidden during training. In this paper, we focus on characterizing and mitigating learning imbalances, i.e., differences in the errors occurring when classifying instances of different classes (aka class-specific risks), under MI-PLL. The phenomenon of learning imbalances has been extensively studied in the context of long-tail learning; however, the nature of MI-PLL introduces new challenges. Our contributions are as follows. From a theoretical perspective, we characterize the learning imbalances by deriving class-specific risk bounds that depend upon the function $\sigma$. Our theory reveals that learning imbalances exist in MI-PLL even when the hidden labels are uniformly distributed. On the practical side, we introduce a technique for estimating the marginal of the hidden labels using only MI-PLL data. Then, we introduce algorithms that mitigate imbalances at training- and testing-time, by treating the marginal of the hidden labels as a constraint. The first algorithm relies on a novel linear programming formulation of MI-PLL for pseudo-labeling. The second one adjusts a model's scores based on robust optimal transport. We demonstrate the effectiveness of our techniques using strong neurosymbolic and long-tail learning baselines, discussing also open challenges.

Related papers

• An Unbiased Risk Estimator for Partial Label Learning with Augmented Classes [46.663081214928226]
  We propose an unbiased risk estimator with theoretical guarantees for PLLAC. We provide a theoretical analysis of the estimation error bound of PLLAC. Experiments on benchmark, UCI and real-world datasets demonstrate the effectiveness of the proposed approach.
  arXiv  Detail & Related papers  (2024-09-29T07:36:16Z)
• On Learning Latent Models with Multi-Instance Weak Supervision [57.18649648182171]
  We consider a weakly supervised learning scenario where the supervision signal is generated by a transition function $sigma$ labels associated with multiple input instances. Our problem is met in different fields, including latent structural learning and neuro-symbolic integration.
  arXiv  Detail & Related papers  (2023-06-23T22:05:08Z)
• Towards Understanding Generalization of Macro-AUC in Multi-label Learning [48.015768048227166]
  We characterize the generalization properties of various learning algorithms based on Macro-AUC. We identify a critical factor of the dataset affecting the generalization bounds: emphthe label-wise class imbalance We propose a new (and more general) McDiarmid-type concentration inequality, which may be of independent interest.
  arXiv  Detail & Related papers  (2023-05-09T08:13:48Z)
• Multi-Instance Partial-Label Learning: Towards Exploiting Dual Inexact Supervision [53.530957567507365]
  In some real-world tasks, each training sample is associated with a candidate label set that contains one ground-truth label and some false positive labels. In this paper, we formalize such problems as multi-instance partial-label learning (MIPL) Existing multi-instance learning algorithms and partial-label learning algorithms are suboptimal for solving MIPL problems.
  arXiv  Detail & Related papers  (2022-12-18T03:28:51Z)
• On Causality in Domain Adaptation and Semi-Supervised Learning: an Information-Theoretic Analysis for Parametric Models [40.97750409326622]
  We study the learning performance of prediction in the target domain from an information-theoretic perspective. We show that in causal learning, the excess risk depends on the size of the source sample at a rate of $O(frac1m)$ only if the labelling distribution between the source and target domains remains unchanged. In anti-causal learning, we show that the unlabelled data dominate the performance at a rate of typically $O(frac1n)$.
  arXiv  Detail & Related papers  (2022-05-10T03:18:48Z)
• Learning from Label Proportions by Learning with Label Noise [30.7933303912474]
  Learning from label proportions (LLP) is a weakly supervised classification problem where data points are grouped into bags. We provide a theoretically grounded approach to LLP based on a reduction to learning with label noise. Our approach demonstrates improved empirical performance in deep learning scenarios across multiple datasets and architectures.
  arXiv  Detail & Related papers  (2022-03-04T18:52:21Z)
• Self-supervised Learning is More Robust to Dataset Imbalance [65.84339596595383]
  We investigate self-supervised learning under dataset imbalance. Off-the-shelf self-supervised representations are already more robust to class imbalance than supervised representations. We devise a re-weighted regularization technique that consistently improves the SSL representation quality on imbalanced datasets.
  arXiv  Detail & Related papers  (2021-10-11T06:29:56Z)
• Active Learning under Label Shift [80.65643075952639]
  We introduce a "medial distribution" to incorporate a tradeoff between importance and class-balanced sampling. We prove sample complexity and generalization guarantees for Mediated Active Learning under Label Shift (MALLS) We empirically demonstrate MALLS scales to high-dimensional datasets and can reduce the sample complexity of active learning by 60% in deep active learning tasks.
  arXiv  Detail & Related papers  (2020-07-16T17:30:02Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.