Name Hidden Layer Regularization
Intent
Add regularization terms that involve activations from inner layers.
Motivation
How can we influence the structure of the model patterns of the inner hidden layers?
References
http://arxiv.org/abs/1507.02672 Semi-Supervised Learning with Ladder Networks
We combine supervised learning with unsupervised learning in deep neural networks. The proposed model is trained to simultaneously minimize the sum of supervised and unsupervised cost functions by backpropagation, avoiding the need for layer-wise pre-training. Our work builds on the Ladder network proposed by Valpola (2015), which we extend by combining the model with supervision.
Note: A regularization term is added to minimize the output of a layer with its corresponding decoder.
http://arxiv.org/pdf/1607.00485v1.pdf Group Sparse Regularization for Deep Neural Networks
We show that a sparse version of the group Lasso penalty is able to achieve competitive performances, while at the same time resulting in extremely compact networks with a smaller number of input features.
http://arxiv.org/abs/1607.02397v1 Enlightening Deep Neural Networks with Knowledge of Confounding Factors
We incorporate information on prominent auxiliary explanatory factors of the data population into existing architectures as secondary objective/loss blocks that take inputs from hidden layers during training.
http://www.ee.cuhk.edu.hk/~xgwang/papers/ouyangWiccv13.pdf Joint Deep Learning for Pedestrian Detection
This paper proposes that they should be jointly learned in order to maximize their strengths through cooperation. We formulate these four components into a joint deep learning framework and propose a new deep network architecture.
http://arxiv.org/pdf/1412.7854v2.pdf Joint Deep Learning for Car Detection
https://arxiv.org/abs/1611.05134 Cost-Sensitive Deep Learning with Layer-Wise Cost Estimation
we propose a novel framework that can be applied to deep neural networks with any structure to facilitate their learning of meaningful representations for cost-sensitive classification problems. Furthermore, the framework allows end- to-end training of deeper networks directly. The framework is designed by augmenting auxiliary neurons to the output of each hidden layer for layer-wise cost estimation, and including the total estimation loss within the optimization objective.
CSDNN starts with a regular DNN with fully-connected layers, but replaces the softmax layer at the end of the DNN by a cost-estimation layer. Each of the K neurons in the cost-estimation layer provides per-class cost estimation with regression instead of per-class probability estimation.