Selective dropout for deep neural networks.
Barrow, Erik; Eastwood, Mark; Jayne, Chrisina
Dropout has been proven to be an effective method for reducing overfitting in deep artificial neural networks. We present 3 new alternative methods for performing dropout on a deep neural network which improves the effectiveness of the dropout method over the same training period. These methods select neurons to be dropped through statistical values calculated using a neurons change in weight, the average size of a neuron's weights, and the output variance of a neuron. We found that increasing the probability of dropping neurons with smaller values of these statistics and decreasing the probability of those with larger statistics gave an improved result in training over 10,000 epochs. The most effective of these was found to be the Output Variance method, giving an average improvement of 1.17 % accuracy over traditional dropout methods.
BARROW, E., EASTWOOD, M. and JAYNE, C. 2016. Selective dropout for deep neural networks. In Hirose, A., Ozawa, S., Doya, K., Ikeda, K., Lee, M. and Liu, D. (eds.) Neural information processing: Proceedings of the 23rd International conference on neural information processing (ICONIP 2016), 16-21 October 2016, Kyoto, Japan. Lecture notes in computer science, 9949. Cham: Springer [online], pages 519-528. Available from: https://doi.org/10.1007/978-3-319-46675-0_57
|Conference Name||23rd International conference on neural information processing (ICONIP 2016)|
|Conference Location||Kyoto, Japan|
|Start Date||Oct 16, 2016|
|End Date||Oct 21, 2016|
|Acceptance Date||Jun 4, 2016|
|Online Publication Date||Sep 29, 2016|
|Publication Date||Sep 29, 2016|
|Deposit Date||Feb 10, 2017|
|Publicly Available Date||Sep 30, 2017|
|Series Title||Lecture notes in computer science|
|Keywords||MNIST; Artificial neural network; Deep learning; Dropout network; Nonrandom dropout; Selective dropout|
BARROW 2016 Selective dropout for deep neural networks
Publisher Licence URL
You might also like
Multiple fake classes GAN for data augmentation in face image dataset.
Video tampering localisation using features learned from authentic content.
Overlap-based undersampling for improving imbalanced data classification.
Toward video tampering exposure: inferring compression parameters from pixels.
Deep imitation learning with memory for robocup soccer simulation.