Learning-hierarchical-invariant

  deep learning一般是学习的层次结构，它也是有一定理论依据的，即模拟人脑的大脑皮层工作，因为大脑皮层的视觉区域也是分层次工作的，越底层的视觉皮层对那些底层特征就越敏感。综上所述，feature learning有这么多应用需求的驱动和生物神经理论上的支持，注定它能够在AI领域中发挥一定的作用。一些实验表明，有些feature learning学习到的特征几乎比所有其它的特征效果要好，比如本文中的ISA模型就是其中一个。(deep learning hierarchy of learning, it is also a certain theoretical basis that simulate the human brain' s cerebral cortex, visual areas of the cerebral cortex is hierarchical work, the more the bottom of the visual cortex is more sensitive to those underlying characteristics . In summary, feature learning so many application needs the support of the drive and biological neural theoretically doomed it to be able to play a certain role in the field of AI. Some experiments show that some feature learning learning to feature almost better than all the other features of the effect of the ISA model such as this is one.)