Deep adaptive networks for image classification
Shusen Zhou, Qingcai Chen, Xiaolong Wang · 2010
This paper proposes a novel classifier called Deep Adaptive Networks (DAN) with deep architecture for image classification. First, we construct a deep and directed belief nets using a set of Restricted Boltzmann Machines (RBM) via greedy and layer-wise unsupervised learning. Then, we refine the parameter space of the deep architecture to adapt the classification demand using global gradient-descent based supervised learning. An exponential loss function is utilized to maximize the separability. Experiments on two real-world image datasets show that the proposed classifier outperforms the representative classification techniques and the existing deep learning methods.