A Computational Perspective on Visual Attention

AttentionBook.md

A Computational Perspective on Visual Attention

Preface

Attention 竟然也会有 起落：70 年代中期的时候，由于计算资源极其有限，Attention 作为一个可以聚焦、节省资源的方式得到了很大的关注；但到了 80 年代，Attention 变得很冷了，因为人们不相信 Attention。现在 Attention 变得更受关注了。本书出版于 2011 年，作者肯定想不到在深度学习统治的现在，Attention 更火了。

这帮认知、AI 交叉的人感兴趣的都是 develop models and systems that are biologically plausible

attention 通常要么被认为是 **selection of a region of interest to guide eye movements **要么被认为是 single-neuron modulation，看作者的意思是这两者是相关的，会有一个 overarching theory of attention 将两者统一起来。

1. Attention—We All Know What It Is

But Do We Really?

Attention 是共同受 sensory 和 intellectual 影响的

Helmholtz （1860）认为 Attention 是 unconscious inference；与之相反，Panum (1858) believed that attention is an activity entirely subservient to an observer’s conscious will

关于解释 Attention 的理论有很多：

1. Early Selection (Broadbent, 1958),
2. using Attenuator Theory (Treisman, 1964)
3. a Late Selection process (Deutsch & Deutsch, 1963; MacKay, 1973; Moray, 1969; Norman, 1968)
4. a two-part process, preattentive followed by attentive processing (Neisser, 1967)
5. as a result of neural synchrony (Milner, 1974), using the metaphor of a spotlight (Shulman, Remington, & McLean, 1979)
6. within Feature Integration Theory (Treisman & Gelade, 1980),
7. as an object-based phenomenon (Duncan, 1984),
8. as a shrink-wrap process (Moran & Desimone, 1985),
9. using the Zoom Lens metaphor (Eriksen & St. James, 1986),
10. as a Premotor Theory subserving eye movements (Rizzolatti, Riggio, Dascola, & Umilta, 1987),
11. as Guided Search (Wolfe, Cave, & Franzel, 1989),
12. as Biased Competition (an extension of the shrink-wrap interpretation; Desimone & Duncan, 1995),
13. as Feature Similarity Gain (Treue & Martinez-Trujillo, 1999)