Generative Adversarial Active Learning

By Jia-Jie Zhu, José Bento. 2017

205 citations recorded by the source corpus.

We propose a new active learning by query synthesis approach using Generative Adversarial Networks (GAN). Different from regular active learning, the resulting algorithm adaptively synthesizes training instances for querying to increase learning speed. We generate queries according to the uncertainty principle, but our idea can work with other active learning principles. We report results from various numerical experiments to demonstrate the effectiveness the proposed approach. In some settings, the proposed algorithm outperforms traditional pool-based approaches. To the best our knowledge, this is the first active learning work using GAN.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Scalable Recollections for Continual Lifelong Learning (2017, methodology): “We also tried an active learning formulation proposed for GANs to learn the best latent code to sample (Zhu & Bento, 2017) at a given time.”
A Wholistic View of Continual Learning with Deep Neural Networks: Forgotten Lessons and the Bridge to Active and Open World Learning (2020, background): “Instead of querying from an unlabelled pool directly, generative adversarial active learning (GAAL) (Zhu and Bento, 2017) and “eﬃcient active learning using conditional generative adversarial network” (Eﬃcient cGAN AL) (Mahapatra et al., 2018) both train GANs to synthesize and label queries.”
Learning to Make Analogies by Contrasting Abstract Relational Structure (2019, background): “LBC also shares similarities to recent generative adversarial active learning approaches (Zhu & Bento, 2017).”
Variational Adversarial Active Learning (2019, background): “Query-synthesizing approaches use generative models to generate informative samples [34, 36, 58] whereas poolbased algorithms use different sampling strategies to determine how to select the most informative samples.”
A Survey of Deep Active Learning (2020, methodology): “For example, GAAL (Generative Adversarial Active Learning) [259] introduced the GAN to the AL query method for the first time.”
A Review on Generative Adversarial Networks: Algorithms, Theory, and Applications (2020, methodology): “GANs have also been used in other machine learning tasks [290] such as active learning [291], [292], online learning [293], ensemble learning [294], zero-shot learning [295], [296], and multi-task learning [297].”
Generative Adversarial Active Learning for Unsupervised Outlier Detection (2018, methodology): “The optimization process can be formulated as: In order to solve the lack of information caused by the ”curse of dimensionality”, we apply the generative adversarial learning framework to outlier detection, which actually performs an active learning process in the proposed method SO-GAAL [26].”
Curriculum Labeling: Revisiting Pseudo-Labeling for Semi-Supervised Learning (2020, methodology): “ing to the output of models classiﬁers. In [9] the authors integrate the active learning framework into the label spreading algorithm which is a graph-based SSL approach for better data selection. In [57] the authors use Generative Adversarial Networks to synthesize training instances for querying t”
Discriminative Active Learning (2019, background): “It should be noted that there have been direct attempts to use GANs for active learning (Zhu & Bento, 2017) in the membership synthesis active learning setting (Settles, 2012), but the results so far have been preliminary and do not improve upon random sampling when evaluating the method on real…”