逐步总结（有待补充）
无监督学习知识框架：
- 这种分类不合适，稀疏编码等也可以从统计学角度看做模型学习与参数选择。实际上，稀疏编码是从1维信号发展起来的表示方法。
- 近年来，稀疏编码逐渐引入信号的先验信息，由非模型向基于模型的转变，学习特色字典。
单层(Sparse,字典-【特征】与ICA字典-【滤波器】代表意思不同)、两层多层训练：
- For example, if k×k image patches are sampled from a set of natural images, an edge at a given orientation may appear at any location, forcing local models to allocate multiple dictionary elements to represent a single underlying orientation. By contrast, a convolutional model only needs to record the oriented structure once, since dictionary elements can be used at all locations.
- The convolutional dictionary does not waste resources modeling similar fi
