Stretching Deep Architectures: A Deep Learning Method without Back-Propagation Optimization

In recent years, researchers have proposed many deep learning algorithms for data representation learning. However, most networks require extensive training and a lot of time to obtain good results. this paper, we propose novel method based on stretching architectures that are composed stacked feature models. Hence, the is called “stretching architectures” (SDA). feedforward propagation SDA, models firstly learned layer by layer, then technique applied map last features high-dimensional space. Since optimized effectively, can be easily calculated, SDA very fast. More importantly, does not need back-propagation optimization, which quite different from existing We tested in visual texture perception, handwritten text recognition, natural image classification applications. Extensive experiments demonstrate advantages over traditional related