DropSample: A New Training Method to Enhance Deep Convolutional Neural Networks for Large-Scale Unconstrained Handwritten Chinese Character Recognition.
Weixin YangLianwen JinDacheng TaoZecheng XieZiyong FengPublished in: CoRR (2015)
Keyphrases
- synthetic data
- significant improvement
- dynamic programming
- high precision
- experimental evaluation
- computational cost
- training process
- training samples
- training algorithm
- classification method
- pairwise
- detection method
- real world
- error rate
- clustering method
- input data
- cost function
- neural network
- supervised learning
- probabilistic model
- multiresolution
- computational complexity
- multiscale
- decision trees