We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: ORDisCo: Effective and Efficient Usage of Incremental Unlabeled Data for Semi-supervised Continual Learning

Abstract: Continual learning usually assumes the incoming data are fully labeled, which might not be applicable in real applications. In this work, we consider semi-supervised continual learning (SSCL) that incrementally learns from partially labeled data. Observing that existing continual learning methods lack the ability to continually exploit the unlabeled data, we propose deep Online Replay with Discriminator Consistency (ORDisCo) to interdependently learn a classifier with a conditional generative adversarial network (GAN), which continually passes the learned data distribution to the classifier. In particular, ORDisCo replays data sampled from the conditional generator to the classifier in an online manner, exploiting unlabeled data in a time- and storage-efficient way. Further, to explicitly overcome the catastrophic forgetting of unlabeled data, we selectively stabilize parameters of the discriminator that are important for discriminating the pairs of old unlabeled data and their pseudo-labels predicted by the classifier. We extensively evaluate ORDisCo on various semi-supervised learning benchmark datasets for SSCL, and show that ORDisCo achieves significant performance improvement on SVHN, CIFAR10 and Tiny-ImageNet, compared to strong baselines.
Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
Cite as: arXiv:2101.00407 [cs.LG]
  (or arXiv:2101.00407v2 [cs.LG] for this version)

Submission history

From: Liyuan Wang [view email]
[v1] Sat, 2 Jan 2021 09:04:14 GMT (22568kb,D)
[v2] Fri, 9 Apr 2021 01:57:03 GMT (22566kb,D)

Link back to: arXiv, form interface, contact.