An Ensemble of Epoch-Wise Empirical Bayes for Few-Shot Learning

@inproceedings{Liu2020AnEO,
  title={An Ensemble of Epoch-Wise Empirical Bayes for Few-Shot Learning},
  author={Yaoyao Liu and B. Schiele and Qianru Sun},
  booktitle={ECCV},
  year={2020}
}
  • Yaoyao Liu, B. Schiele, Qianru Sun
  • Published in ECCV 2020
  • Computer Science, Mathematics
  • Few-shot learning aims to train efficient predictive models with a few examples. The lack of training data leads to poor models that perform high-variance or low-confidence predictions. In this paper, we propose to meta-learn the ensemble of epoch-wise empirical Bayes models (E3BM) to achieve robust predictions. "Epoch-wise" means that each training epoch has a Bayes model whose parameters are specifically learned and deployed. "Empirical" means that the hyperparameters, e.g., used for learning… CONTINUE READING
    8 Citations

    Figures and Tables from this paper

    Free Lunch for Few-shot Learning: Distribution Calibration
    • PDF
    Boosting Few-Shot Learning With Adaptive Margin Loss
    • 9
    • PDF
    Attentive Graph Neural Networks for Few-Shot Learning
    • 1
    • PDF
    Comparative Analysis on Classical Meta-Metric Models for Few-Shot Learning
    • PDF
    Interventional Few-Shot Learning
    • 4
    • PDF

    References

    SHOWING 1-10 OF 96 REFERENCES
    Optimization as a Model for Few-Shot Learning
    • 1,358
    Diversity With Cooperation: Ensemble Methods for Few-Shot Classification
    • 52
    • Highly Influential
    • PDF
    Learning to Self-Train for Semi-Supervised Few-Shot Classification
    • 52
    • PDF
    TADAM: Task dependent adaptive metric for improved few-shot learning
    • 360
    • Highly Influential
    • PDF
    Meta-Learning for Semi-Supervised Few-Shot Classification
    • 337
    • PDF
    Meta-Transfer Learning through Hard Tasks
    • 3
    • PDF
    Empirical Bayes Meta-Learning with Synthetic Gradients
    • 1
    • Highly Influential
    • PDF
    Meta-Learning With Differentiable Convex Optimization
    • 248
    • PDF
    Probabilistic Model-Agnostic Meta-Learning
    • 226
    • Highly Influential
    • PDF
    Learning to Propagate Labels: Transductive Propagation Network for Few-Shot Learning
    • 159
    • PDF