| 2021 |
Bayesian Active Learning by Soft Mean Objective Cost of Uncertainty |
Zhao et al. |
AISTATS |
- |
Clssification |
Uncertainty, BNNs, None, Tra, Hard |
UCI User Knowledge dataset, center dataset |
these methods are not guaranteed to converge to the optimal classifier of the true model because MOCU is not strictly concave. |
| 2021 |
Active Learning with Maximum Margin Sparse Gaussian Processes |
Shi and Yu |
AISTATS |
- |
Multi-class classification |
maximum-margin, Gaussian Process,None, Tra, Hard |
generate a 2D synthetic dataset, Dermatology I, Dermatology II, Yeast, Penstroke, Auto-Drive, Reuters |
|
| 2021 |
Revisiting Model-Agnostic Private Learning: Faster Rates and Active Learning |
Liu et al. |
AISTATS |
- |
|
Ensembles, Many Classifiers, None, Tra, Hard |
|
We fill in this gap by introducing the Tsybakov Noise Condition (TNC) and establish stronger and more interpretable learning bounds. |
| 2021 |
Feedback Coding for Active Learning |
Canal et al. |
AISTATS |
code |
Bayesian logistic regression |
Posterior Matching, BNNs, None, Tra, Hard |
UCI dataset, vehicle, letter, austra, and wdbc. |
Information Theory |
| 2021 |
Towards Understanding the Behaviors of Optimal Deep Active Learning Algorithms |
Zhou et al. |
AISTATS |
code |
Object/Intent Classification, Named Entity Recognition |
Any |
Fashion-MNIST, |
optimal data acquisition order |
| 2021 |
Active Learning under Label Shift |
Zhao et al. |
AISTATS |
- |
Streaming |
Hybrid, BNNs, None, Tra, Hard |
|
We address the problem of active learning under label shift: when the class proportions of source and target domains differ. |