We present a novel approach for training a multi-layered perceptron (MLP) in a semi-supervised fashion. Our objective function, when optimized, balances training set accuracy with fidelity to a graph-based manifold over all points. Additionally, the objective favors smoothness via an entropy regularizer over classifier outputs as well as straightforward 2 regularization. Our approach also scales well enough to enable large-scale training. The results demonstrate significant improvement on several phone classification tasks over baseline MLPs.
Cite as: Malkin, J., Subramanya, A., Bilmes, J. (2009) On the semi-supervised learning of multi-layered perceptrons. Proc. Interspeech 2009, 660-663, doi: 10.21437/Interspeech.2009-231
@inproceedings{malkin09_interspeech,
author={Jonathan Malkin and Amarnag Subramanya and Jeff Bilmes},
title={{On the semi-supervised learning of multi-layered perceptrons}},
year=2009,
booktitle={Proc. Interspeech 2009},
pages={660--663},
doi={10.21437/Interspeech.2009-231}
}