English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Unsupervised pre-training was a critical technique for training deep neuralnetworks years ago. With sufficient labeled data and modern trainingtechniques, it is possible to train very deep neural networks from scratch in apurely supervised manner nowadays. However, unlabeled data is easier to obtainand usually of very large scale. How to make use of them better to helpsupervised learning is still a well-valued topic. In this paper, we investigateconvolutional denoising auto-encoders to show that unsupervised pre-trainingcan still improve the performance of high-level image related tasks such asimage classification and semantic segmentation. The architecture we use is aconvolutional auto-encoder network with symmetric shortcut connections. Weempirically show that symmetric shortcut connections are very important forlearning abstract representations via image reconstruction. When no extraunlabeled data are available, unsupervised pre-training with our network canregularize the supervised training and therefore lead to better generalizationperformance. With the help of unsupervised pre-training, our method achievesvery competitive results in image classification using very simpleall-convolution networks. When labeled data are limited but extra unlabeleddata are available, our method achieves good results in several semi-supervisedlearning tasks.

Learning Deep Representations Using Convolutional Auto-encoders with Symmetric Skip Connections