deepGenerativeModels

From statwiki
Jump to: navigation, search

Problem

To properly train a neural network a large labeled dataset, however large datasets of labeled data are not always available. This paper covers a method of generating labeled data for neural network training.


Methodology

Future Work

These models scale linearly in proportion to the number of classes in the data sets. The number of evaluations could be reduced by truncating the posterior mass or by using error-correcting output codes.