【人工智能原理】Natural Image Statistics A probabilistic approach to early computational vision

目录

Introduction
Background
Statistics of linear features
Nonlinear features & dependency of linear features
Time, colour and stereo
Conclusion

信息论构建了生物感知系统与环境特征的联系

motivated by developments in information theory, Attneave (1954) suggested that the goal of visual perception is to produce an efficient representation of the incoming signal.

人类视觉神经学

视觉信号的传输
视网膜-视神经-LGN(颞下皮质)-V1(初级视觉皮层)-V2-V4

卷积网络描述了V1的三种性质：

0. 空间映射
1. 简单细胞，线性函数
2. 复杂细胞，池化单元，maxout单元

未解问题

1. 目前的卷积网络纯是视觉的，听觉呢？
2. 人对场景的理解，如空间几何关系?
3. 人类视觉的反馈机制?
4. 反向相关实验

先验

coding, estimation, and decision theories all rely heavily on a statistical “prior” model of the environment.

Ecological adaptation provides prior information
Generative models
Projection onto the retina loses information