Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
Обучение представлений без учителя с помощью глубоких сверточных генеративных состязательных сетей
2015-11-19
SCID: 54.1/7rkf2qat
Discuss with AI
DCGANconvolutional neural networksdeep convolutional generative adversarial networkslearned image featuresunsupervised representation learning
Figures from the paper
Abstract (AI)
In recent years, supervised learning with convolutional networks (CNNs) has seen huge adoption in computer vision applications. Comparatively, unsupervised learning with CNNs has received less attention. In this work we hope to help bridge the gap between the success of CNNs for supervised learning and unsupervised learning. We introduce a class of CNNs called deep convolutional generative adversarial networks (DCGANs), that have certain architectural constraints, and demonstrate that they are a strong candidate for unsupervised learning. Training on various image datasets, we show convincing evidence that our deep convolutional adversarial pair learns a hierarchy of representations from object parts to scenes in both the generator and discriminator. Additionally, we use the learned features for novel tasks - demonstrating their applicability as general image representations.
Key Findings
1
DCGANs learn a hierarchy of representations from object parts to scenes in both the generator and discriminator when trained on image datasets.
2
DCGANs provide strong empirical evidence that convolutional architectures can be effective for unsupervised representation learning.
3
Introduced deep convolutional generative adversarial networks (DCGANs), a class of CNNs with specific architectural constraints for unsupervised learning.
4
Learned DCGAN features are applicable as general image representations for novel tasks, demonstrating transferability beyond training.
Research Object
Deep convolutional generative adversarial networks (DCGANs) trained on image datasets
Research Subject
Unsupervised representation learning ability of the DCGANs, i.e., the hierarchy of learned image features in generator and discriminator and their applicability as general image representations
Publication Details
Publication Date
2015-11-19
Journal
Publisher
ISSN
Cited by
12235
Open access PDF
Access Type
Author Information
Download PDF
Subscribe to digest
References available in scid.ai2
Cited by20
Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks2017
Image-to-Image Translation with Conditional Adversarial Networks2017
A survey on Image Data Augmentation for Deep Learning2019
Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network2017
Least Squares Generative Adversarial Networks2017
A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects2021
Generative Adversarial Networks: An Overview2018
Automatic Chemical Design Using a Data-Driven Continuous Representation of Molecules2018
Self-Attention Generative Adversarial Networks2018
Understanding and Mitigating Gradient Flow Pathologies in Physics-Informed Neural Networks2021
Federated Learning in Mobile Edge Networks: A Comprehensive Survey2020
Review of Deep Learning Algorithms and Architectures2019
Artificial intelligence and smart vision for building and construction 4.0: Machine and deep learning methods and applications2022
A Survey of Deep Learning Methods for Cyber Security2019
Deep Learning for Anomaly Detection in Time-Series Data: Review, Analysis, and Guidelines2021
Reconstruction of three-dimensional porous media using generative adversarial neural networks2017
Machine Learning Methods for Small Data Challenges in Molecular Science2023
MuseGAN: Multi-track Sequential Generative Adversarial Networks for Symbolic Music Generation and Accompaniment2018
A survey on data‐efficient algorithms in big data era2021
Machine Learning Seismic Wave Discrimination: Application to Earthquake Early Warning2018