• Mathematics, Computer Science
  • Published in NeurIPS 2019

BatchBALD: Efficient and Diverse Batch Acquisition for Deep Bayesian Active Learning

@inproceedings{Kirsch2019BatchBALDEA,
  title={BatchBALD: Efficient and Diverse Batch Acquisition for Deep Bayesian Active Learning},
  author={Andreas Kirsch and Joost R. van Amersfoort and Yarin Gal},
  booktitle={NeurIPS},
  year={2019}
}
We develop BatchBALD, a tractable approximation to the mutual information between a batch of points and model parameters, which we use as an acquisition function to select multiple informative points jointly for the task of deep Bayesian active learning. [...] Key Method We compare BatchBALD to the commonly used approach for batch data acquisition and find that the current approach acquires similar and redundant points, sometimes performing worse than randomly acquiring data. We finish by showing that, using…Expand Abstract

Citations

Publications citing this paper.

References

Publications referenced by this paper.
SHOWING 1-10 OF 36 REFERENCES

CINIC-10 is not ImageNet or CIFAR-10

VIEW 3 EXCERPTS
HIGHLY INFLUENTIAL

Deep Bayesian Active Learning with Image Data

VIEW 6 EXCERPTS

EMNIST: Extending MNIST to handwritten letters

VIEW 3 EXCERPTS
HIGHLY INFLUENTIAL

Adam: A Method for Stochastic Optimization

VIEW 2 EXCERPTS
HIGHLY INFLUENTIAL

A new outlook of Shannon's information measures

VIEW 3 EXCERPTS
HIGHLY INFLUENTIAL

Variational Adversarial Active Learning

VIEW 2 EXCERPTS