Abstract
In recent years, supervised learning with convolutional networks (CNNs) has\nseen huge adoption in computer vision applications. Comparatively, unsupervised\nlearning with CNNs has received less attention. In this work we hope to help\nbridge the gap between the success of CNNs for supervised learning and\nunsupervised learning. We introduce a class of CNNs called deep convolutional\ngenerative adversarial networks (DCGANs), that have certain architectural\nconstraints, and demonstrate that they are a strong candidate for unsupervised\nlearning. Training on various image datasets, we show convincing evidence that\nour deep convolutional adversarial pair learns a hierarchy of representations\nfrom object parts to scenes in both the generator and discriminator.\nAdditionally, we use the learned features for novel tasks - demonstrating their\napplicability as general image representations.\n
Keywords
Artificial intelligenceComputer scienceConvolutional neural networkUnsupervised learningDeep learningDiscriminatorFeature learningRepresentation (politics)Adversarial systemGenerative grammarMachine learningGenerator (circuit theory)Class (philosophy)HierarchyPattern recognition (psychology)
Affiliated Institutions
Related Publications
Extracting and composing robust features with denoising autoencoders
Previous work has shown that the difficulties in learning deep generative or discriminative models can be overcome by an initial unsupervised learning step that maps inputs to u...
Sparse Feature Learning for Deep Belief Networks
Unsupervised learning algorithms aim to discover the structure hidden in the data, and to learn representations that are more suitable as input to a supervised machine than the ...
A State-of-the-Art Survey on Deep Learning Theory and Architectures
In recent years, deep learning has garnered tremendous success in a variety of application domains. This new field of machine learning has been growing rapidly and has been appl...
Greedy Layer-Wise Training of Deep Networks
Complexity theory of circuits strongly suggests that deep architectures can be much more efficient (sometimes exponentially) than shallow architectures, in terms of computationa...
Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences
Significance Learning biological properties from sequence data is a logical step toward generative and predictive artificial intelligence for biology. Here, we propose scaling a...
Publication Info
- Year
- 2015
- Type
- preprint
- Citations
- 7618
- Access
- Closed
External Links
Social Impact
Altmetric
PlumX Metrics
Social media, news, blog, policy document mentions
Citation Metrics
7618
OpenAlex
Cite This
Alec Radford,
Luke Metz,
Soumith Chintala
(2015).
Unsupervised Representation Learning with Deep Convolutional Generative\n Adversarial Networks.
arXiv (Cornell University)
.
https://doi.org/10.48550/arxiv.1511.06434
Identifiers
- DOI
- 10.48550/arxiv.1511.06434