At each acquisition step, we score a batch of candidate unlabeled data samples based on the acquisition function used and the deep model trained on the labelled set.

Batch Mode DeepAL

It combines BCNNs (Bayesian Convolutional Neural Networks) with AL methods
to adapt BALD to the deep learning environment, thereby developing a new AL framework for high-dimensional data.

Deep Bayesian Active Learning

Deep Bayesian Active Learning with Image Data

Bayesian Active Learning for Classification and Preference Learning

DBAL is unsuitable for large datasets due to the need for batch sampling.  DBAL allows the use of dropout in testing for better confidence estimation. the performance of this method is similar to the performance of using neural network SR as uncertainty sampling, which requires vigilance.

Why DBAL is not Suitable for Large Datasets?

Deep Ensemble Bayesian Active Learning

Task-Aware Variational Adversarial Active Learning

Variational Adversarial Active Learning

Structure Comparison Chart of VAAL and TA-VAAL

1. Batch Mode DeepAL.
2. Uncertainty-based and Hybrid Query Strategies.
3. Deep Bayesian Active Learning. (DBAL)
4. Density-based Methods.
5. Automated Design of DeepAL.

University of Michigan - Ann Arbor

Active learning (AL) attempts to maximize a model’s performance gain while annotating the fewest samples possible.

Active Learning

A Survey of Deep Active Learning

AL approaches can be divided into membership query synthesis, stream-based selective sampling, and pool-based AL from application scenarios.

Active Learning Categories

1. Uncertainty-based Strategy.
2. Diversity-based Strategy.
3. Expected Model Change.

Query Strategy Categories of Active Learning

1. Model uncertainty in Deep Learning. 
2. Insufficient data for labeled samples.
3. Processing pipeline inconsistency.

Challenges of Combining Deep Learning and Active Learning

We define $U^n = {X,Y}$ as an unlabeled dataset with n labels. 
$X$ is the sample space. 
$Y$ is the label space.
$P(x,y)$ is a potential distribution.
$L^m = {X,Y}$ is the current labeled training set with m samples.
Query strategy is $Q$ which maps $U^n$ to $L^m$.


Learn Before

Related

Learn After