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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Classifiers for the semi-supervised setting often combine strong supervisedmodels with additional learning objectives to make use of unlabeled data. Thisresults in powerful though very complex models that are hard to train and thatdemand additional labels for optimal parameter tuning, which are often notgiven when labeled data is very sparse. We here study a minimalisticmulti-layer generative neural network for semi-supervised learning in a formand setting as similar to standard discriminative networks as possible. Basedon normalized Poisson mixtures, we derive compact and local learning and neuralactivation rules. Learning and inference in the network can be scaled usingstandard deep learning tools for parallelized GPU implementation. With thesingle objective of likelihood optimization, both labeled and unlabeled dataare naturally incorporated into learning. Empirical evaluations on standardbenchmarks show, that for datasets with few labels the derived minimalisticnetwork improves on all classical deep learning approaches and is competitivewith their recent variants without the need of additional labels for parametertuning. Furthermore, we find that the studied network is the best performingmonolithic (`non-hybrid') system for few labels, and that it can be applied inthe limit of very few labels, where no other system has been reported tooperate so far.

Neural Simpletrons - Minimalistic Directed Generative Networks for Learning with Few Labels