GoPiGo 小汽車︰格點圖像算術《色彩空間》故事

所有的故事都有一個開始,『知覺量化』的故事也是一樣︰

一六一四年 John Napier 約翰‧納皮爾在一本名為《 Mirifici Logarithmorum Canonis Descriptio  》── 奇妙的對數規律的描述 ── 的書中,用了三十七頁解釋『對數log ,以及給了長達九十頁的對數表。這有什麼重要的嗎?想一想即使在今天用『鉛筆』和『紙』做大位數的加減乘除,尚且困難也很容易算錯,就可以知道對數的發明,對計算一事貢獻之大的了。如果用一對一對應的觀點來看,對數把『乘除』運算『變換加減』運算

\log {a * b} = \log{a} + \log{b}

\log {a / b} = \log{a} - \log{b}

,更不要說還可以算『平方』、『立方』種種和開『平方根』、『立方根』等等的計算了。

\log {a^n} = n * \log{a}

傳聞納皮爾還發明了的『骨頭計算器』,他的書對於之後的天文學、力學、物理學、占星學的發展都有很大的影響。他的運算變換 Transform 的想法,開啟了『換個空間解決數學問題』的大門,比方『常微分方程式的  Laplace Transform』與『頻譜分析的傅立葉變換』等等。

這個對數畫起來是這個樣子︰

Rendered by QuickLaTeX.com

不只如此這個對數關係竟然還跟人類之『五官』── 眼耳鼻舌身 ── 受到『刺激』── 色聲香味觸 ── 的『感覺』強弱大小有關。一七九五年出生的 Ernst Heinrich Weber 韋伯,一位德國物理學家,是一位心理物理學的先驅,他提出感覺之『方可分辨』JND just-noticeable difference 的特性。比方說你提了五公斤的水,再加上半公斤,可能感覺差不了多少,要是你沒提水,說不定會覺的突然拿著半公斤的水很重。也就是說在『既定的刺激』下, 感覺的方可分辨性大小並不相同。韋伯實驗後歸結成一個關係式︰

ΔR/R = K

R:  既有刺激之物理量數值
ΔR:  方可分辨 JND 所需增加的刺激之物理量數值
K: 特定感官之常數,不同的感官不同

。之後  Gustav Theodor Fechner  費希納,一位韋伯派的學者,提出『知覺』perception 『連續性假設,將韋伯關係式改寫為︰

dP = k  \frac {dS}{S}

,求解微分方程式得到︰

P = k \ln S + C

假如刺激之物理量數值小於 S_0 時,人感覺不到 P = 0,就可將上式寫成︰

P = k \ln \frac {S}{S_0}

這就是知名的韋伯-費希納定律,它講著:在絕對閾限 S_0 之上,主觀知覺之強度的變化與刺激之物理量大小的改變呈現自然對數的關係,也可以說,如果刺激大小按著幾何級數倍增,所引起的感覺強度卻只依造算術級數累加。

─── 摘自《千江有水千江月

 

在故事展開前特此提醒讀者︰

『量』與『數』大不同。『物理量』是有『單位』的。

由於主觀『知覺強度』一般符合『物理量』刺激之『對數關係』。然而『對數』是『非線性』的。所以從『因次分析』的觀點來看,

P = k \ln \frac {S}{S_0} 之『無因次』 \frac{S}{S_0} 『表述』實優於

P = k \ln S + C 也。

追求五官之『絕對 □ 感』者,或當知『等距量表』哩︰

170px-Pakkanen
溫度計
量冷熱

魯班尺
魯班尺
度吉凶

一九四七年,匈牙利之美籍猶太人數學家,現代電腦創始人之一。約翰‧馮‧諾伊曼 Jhon Von Neumann 和德國-美國經濟學家奧斯卡‧摩根斯特恩 Oskar Morgenstern 提出只要『個體』的『喜好性』之『度量』滿足『四條公設』,那麼『個體』之『效用函數』就『存在』,而且除了『零點』的『規定』,以及『等距長度』之『定義』之外,這個『效用函數』還可以說是『唯一』的。就像是『個體』隨身攜帶的『理性』之『溫度計』一樣,能在任何『選擇』下,告知最大『滿意度』與『期望值』。現今這稱之為『期望效用函數理論』 Expected Utility Theory。

由於每個人的『冷熱感受』不同,所以『溫度計』上的『刻度』並不是代表數學上的一般『數字』,通常這一種比較『尺度』只有『差距值』有相對『強弱』意義,『數值比值』並不代表什麼意義,就像說,攝氏二十度不是攝氏十度的兩倍熱。這一類『尺度』在度量中叫做『等距量表』 Interval scale 。

溫度計』量測『溫度』的『高低』,『理性』之『溫度計』度量『選擇』的『優劣』。通常在『實驗經濟學』裡最廣泛採取的是『彩票選擇實驗』 lottery- choice experiments,也就是講,請你在『眾多彩票』中選擇一個你『喜好』 的『彩票』。

這樣就可以將一個有多種『機率p_i,能產生互斥『結果A_i 的『彩票L 表示成︰

L = \sum \limits_{i=1}^{N} p_i A_i ,  \  \sum \limits_{i=1}^{N} p_i  =1,  \ i=1 \cdots N

如此『期望效用函數理論』之『四條公設』可以表示為︰

完整性公設】Completeness

L\prec MM\prec L,或 L \sim M

任意的兩張『彩票』都可以比較『喜好度』 ,它的結果只能是上述三種關係之一,『偏好 ML\prec M,『偏好 LM\prec L,『無差異L \sim M

遞移性公設】 Transitivity

如果 L \preceq M,而且 M \preceq N,那麼 L \preceq N

連續性公設】 Continuity

如果 L \preceq M\preceq N , 那麼存在一個『機率p\in[0,1] ,使得 pL + (1-p)N = M

獨立性公設】 Independence

如果 L\prec M, 那麼對任意的『彩票N 與『機率p\in(0,1],滿足 pL+(1-p)N \prec pM+(1-p)N

對於任何一個滿足上述公設的『理性經紀人』 rational agent ,必然可以『建構』一個『效用函數u,使得 A_i \rightarrow u(A_i),而且對任意兩張『彩票』,如果 L\prec M \Longleftrightarrow \  E(u(L)) < E(u(M))。此處 E(u(L)) 代表對 L彩票』的『效用期望值』,簡記作 Eu(L),符合

Eu(p_1 A_1 + \ldots + p_n A_n) = p_1 u(A_1) + \cdots + p_n u(A_n)

它在『微觀經濟學』、『博弈論』與『決策論』中,今天稱之為『預期效用假說』 Expected utility hypothesis,指在有『風險』的情況下,任何『個體』所應該作出的『理性選擇』就是追求『效用期望值』的『最大化』。假使人生中的『抉擇』真實能夠如是的『簡化』,也許想得到『快樂』與『幸福』的辦法,就清楚明白的多了。然而有人認為這個『假說』不合邏輯。一九五二年,法國總體經濟學家莫里斯‧菲力‧夏爾‧阿萊斯 Maurice Félix Charles Allais ── 一九八八年,諾貝爾經濟學獎的得主 ── 作了一個著名的實驗,看看實際上人到底是怎麼『做選擇』的,這個『阿萊斯』發明的『彩票選擇實驗』就是大名鼎鼎的『阿萊斯悖論』 Allais paradox 。

針對百人測試所設計的『彩票』:

彩票甲:百分之百的機會得到一百萬元。【期望值 100 萬】

彩票乙:百分之十的機會得到五百萬元,百分之八十九的機會得到一百萬元,百分之一的機會什麼也得不到。【期望值 139 萬】

實驗結果:絕大多數人選擇甲而非乙。

然後改用另一組『彩票』,對同一群人繼續進行測試︰

彩票丙:百分之十一的機會得到一百萬元,百分之八十九的機會什麼也得不到。【期望值 11 萬】

彩票丁:百分之十的機會得到五百萬元,百分之九十的機會什麼也得不到。【期望值 50 萬】

實驗結果:絕大多數人選擇丁而非丙。

那麼這又是為什麼呢?也許說設想『人只是理性的』的這種想法,並不符合『合理性』,畢竟『人的心理』是『複雜的』,而且『人類行為』也是『多樣的』。於是自一九七九年起,以色列裔美國心理學家丹尼爾‧卡內曼 Daniel Kahneman 和以色列著名認知心理學者阿摩司‧特沃斯基 Amos Tversky 系統的研究『行為經濟學』 behavioral economic theory 這一領域,開創了現今稱為的『展望理論』prospect theory,試圖回答『為什麼』人是這麼『做選擇』的,此『前景理論』這麼講︰

People make decisions based on the potential value of losses and gains rather than the final outcome, and that people evaluate these losses and gains using certain heuristics.

這一個『描述性』理論認為,每個人基於自身所處之『參考點』之『不同』,面對『風險』就會有不同的『態度』。他們假設一個人的『得失衡量』可以表示成︰

U = \sum \limits_{i=1}^N w(p_i)v(A_i)

,此處 A_i 是各個可能結果,而 v 是『價值函數』 value function ,表示不同可能結果,在決策者心中的『相對價值』。而 w 是『機會權重函數』 probability weighting function ,藉此表現通常人對於『極不可能』發生的事,往往會『過度反應』 over-react,而對『高度可能』出現的事,常常又會『反應不及』 under-react。從而形成一條穿過『參考點』的『S 型曲線』。那個 U 就是一個人在作『得失決策』時的『總體評估』,或者說『預期效用』。

220px-Courbe_niveau.svg

300px-Contour2D.svg

505px-Arctic.svg

Valuefun

價值函數 value function

Loop_isallobaric_tendencies

這條『S 型曲線』的不對稱性呈現出,當人們面對一個『損失』的『結果』,所產生之『厭惡感』或者說『傷感情』,比『獲益』之『情況』下所生的『滿意度』也許講『感覺好』,更為『強烈』。這使『展望理論』基本上不同於『期望效用函數理論』。有人將此理論的引申結論,整理成︰

確定效應:處於穫益狀態時,多數人是風險厭惡者。
反射效應:處於損失狀態時,多數人是風險喜好者。
損失規避:多數人對損失比對穫益敏感。
參照依賴:多數人對得失的判斷往往由參照點決定。

── 『人的行為』應當用著『純理性』來『定義』嗎?

還是應該要講『有人情』真的就『不合理』的嗎??──

─── 摘自《物理哲學·下中…

 

亦可免於誤解所謂『費希納』之個體局部以及『史蒂文斯』的整體大眾『心理學』議論乎︰

Stevens’s power law

Stevens’s power law is a proposed relationship between the magnitude of a physical stimulus and its perceived intensity or strength. It is often considered to supersede the Weber–Fechner law on the basis that it describes a wider range of sensations, although critics argue that the validity of the law is contingent on the virtue of approaches to the measurement of perceived intensity that are employed in relevant experiments. In addition, a distinction has been made between local psychophysics, where stimuli are discriminated only with a certain probability, and global psychophysics, where the stimuli would be discriminated correctly with near certainty (Luce & Krumhansl, 1988). The Weber–Fechner law and methods described by L. L. Thurstone are generally applied in local psychophysics, whereas Stevens’s methods are usually applied in global psychophysics.

The theory is named after psychophysicist Stanley Smith Stevens (1906–1973). Although the idea of a power law had been suggested by 19th-century researchers, Stevens is credited with reviving the law and publishing a body of psychophysical data to support it in 1957.

The general form of the law is

  {\displaystyle \psi (I)=kI^{a},}

where I is the magnitude of the physical stimulus, ψ(I) is the subjective magnitude of the sensation evoked by the stimulus, a is an exponent that depends on the type of stimulation, and k is a proportionality constant that depends on the units used.

The table to the bottom lists the exponents reported by Stevens.

 

Continuum Exponent  a Stimulus condition
Loudness 0.67 Sound pressure of 3000 Hz tone
Vibration 0.95 Amplitude of 60 Hz on finger
Vibration 0.6 Amplitude of 250 Hz on finger
Brightness 0.33 5° target in dark
Brightness 0.5 Point source
Brightness 0.5 Brief flash
Brightness 1 Point source briefly flashed
Lightness 1.2 Reflectance of gray papers
Visual length 1 Projected line
Visual area 0.7 Projected square
Redness (saturation) 1.7 Red–gray mixture
Taste 1.3 Sucrose
Taste 1.4 Salt
Taste 0.8 Saccharin
Smell 0.6 Heptane
Cold 1 Metal contact on arm
Warmth 1.6 Metal contact on arm
Warmth 1.3 Irradiation of skin, small area
Warmth 0.7 Irradiation of skin, large area
Discomfort, cold 1.7 Whole-body irradiation
Discomfort, warm 0.7 Whole-body irradiation
Thermal pain 1 Radiant heat on skin
Tactual roughness 1.5 Rubbing emery cloths
Tactual hardness 0.8 Squeezing rubber
Finger span 1.3 Thickness of blocks
Pressure on palm 1.1 Static force on skin
Muscle force 1.7 Static contractions
Heaviness 1.45 Lifted weights
Viscosity 0.42 Stirring silicone fluids
Electric shock 3.5 Current through fingers
Vocal effort 1.1 Vocal sound pressure
Angular acceleration 1.4 5 s rotation
Duration 1.1 White-noise stimuli

 

只要『批評』

Criticisms

Stevens generally collected magnitude estimation data from multiple observers, averaged the data across subjects, and then fitted a power function to the data. Because the fit was generally reasonable, he concluded the power law was correct. This approach ignores any individual differences that may obtain and indeed it has been reported that the power relationship does not always hold as well when data are considered separately for individual respondents (Green & Luce 1974).

Another issue is that the approach provides neither a direct test of the power law itself nor the underlying assumptions of the magnitude estimation/production method.

Stevens’s main assertion was that using magnitude estimations/productions respondents were able to make judgements on a ratio scale (i.e., if x and y are values on a given ratio scale, then there exists a constant k such that x = ky). In the context of axiomatic psychophysics, (Narens 1996) formulated a testable property capturing the implicit underlying assumption this assertion entailed. Specifically, for two proportions p and q, and three stimuli, x, y, z, if y is judged p times x, z is judged q times y, then t = pq times x should be equal to z. This amounts to assuming that respondents interpret numbers in a veridical way. This property was unambiguously rejected (Ellermeier & Faulhammer 2000, Zimmer 2005). Without assuming veridical interpretation of numbers, (Narens 1996) formulated another property that, if sustained, meant that respondents could make ratio scaled judgments, namely, if y is judged p times x, z is judged q times y, and if y is judged q times x, z is judged p times y, then z should equal z. This property has been sustained in a variety of situations (Ellermeier & Faulhammer 2000, Zimmer 2005).

Because Stevens fit power functions to data, his method did not provide a direct test of the power law itself. (Luce 2002), under the condition that respondents’ numerical distortion function and the psychophysical functions could be separated, formulated a behavioral condition equivalent to the psychophysical function being a power function. This condition was confirmed for just over half the respondents, and the power form was found to be a reasonable approximation for the rest (Steingrimsson & Luce 2006).

It has also been questioned, particularly in terms of signal detection theory, whether any given stimulus is actually associated with a particular and absolute perceived intensity; i.e. one that is independent of contextual factors and conditions. Consistent with this, Luce (1990, p. 73) observed that “by introducing contexts such as background noise in loudness judgements, the shape of the magnitude estimation functions certainly deviates sharply from a power function”.

 

仍在,『理論』尚未『 ○ 滿』耶??