Untangling complex syste.., p.99

Untangling Complex Systems, page 99

 

Untangling Complex Systems
Select Voice:
Brian (uk)
Emma (uk)  
Amy (uk)
Eric (us)
Ivy (us)
Joey (us)
Salli (us)  
Justin (us)
Jennifer (us)  
Kimberly (us)  
Kendra (us)
Russell (au)
Nicole (au)



Larger Font   Reset Font Size   Smaller Font  



  tribution of exponential terms.

  The term Pr( h) is the “prior probability,” and it codifies our knowledge about a possible spectrum

  before collecting experimental data. It is expressed as

  Pr( )

  h

  eS

  ∝ [B.6]

  with S being the Information Entropy:

  S( h) = − h( )log  h(α) d

  ∫ α   α [B.7]

  Finally, the term Pr( h D) is the “posterior probability.” A single answer to the inverse Laplace

  transform problem can be obtained by maximizing Pr( h D). The “posterior probability” is

  S−1 2

  χ

  e 2

  eQ

  Pr( h D) =

  =

  [B.8]

  ZlZD

  ZlZD

  Maximizing Pr( h D) means finding the maximum of the exponent Q = S − (1 2) 2

  / χ . Q is maximized

  when the Entropy S is maximized whereas χ 2 is minimized. MEM becomes a “tug of war” between

  maximizing the Entropy S and minimizing the χ 2 (see Figure B.3).

  Mathematically, the Information Entropy has the important property that when maximizing it,

  no possibilities are excluded. In other words, maximizing S in the absence of the experimental

  data, we impose no correlations on the spectrum α( τ); we choose the spectrum that is minimally

  structured or “maximally noncommittal.” This concept can be illustrated by a funny numerical

  example: the so-called “Kangaroo problem” (see Figure B.4). The information available is that (1/3)

  510

  Appendix B

  n 0

  2

  1

  data datum

  S( h) = ∫ [− h( τ)log h( τ)] dτ

  χ 2 =

  i − fiti

  n 0

  σ

  data i=1

  i

  FIGURE B.3 The Maximum Entropy Method is a “tug of war” between the Information Entropy and the χ 2.

  Left-handed

  Left-handed

  True

  False

  True

  False

  p

  1

  x

  p 1

  x

  True

  True

  −

  3

  x

  Blue-eyed

  p

  1

  1

  2

  p 3

  Blue-eyed

  −

  + x

  False

  False

  3

  x

  3

  FIGURE B.4 The example of blue-eyed and left-handed kangaroos.

  of kangaroos have blue eyes, and (1/3) are left-handed. The question is how many kangaroos are

  both left-handed and blue-eyed. Of course, the two properties are unconnected (as far as we know,

  unless one day we will discover that the genes responsible for blue-eyes are somehow correlated

  with the genes responsible for left-handedness). We have four possibilities shown in the left Table of

  Figure B.4. The normalization imposes that

  px + p 1 + p 2 + p 3 = 1 [B.9]

  We know that

  1 

  px + p 1 =  

   3  [B.10]

  1 

  px + p 2 =  

   3 

  If we fix px = x, p 1, and p 2 will be ( /

  1 3 − x), and p 3 will be ( /

  1 3 + x), as shown in the table on the

  right of Figure B.4. According to the theory, the probability of two independent, uncorrelated events is given by the product of their single probabilities. In our case, it is expected that the probability

  to have left-handed and blue-eyed kangaroos is ( /

  1 )(

  3 /

  1 )

  3 , i.e., ( /

  1 )

  9 . Is there some function of the

  probabilities, pi, which, when maximized, yields the maximally noncommittal result? The answer

  is yes, and it is just the Information Entropy function:

  

   1

    1

    1

    1

  

  S = −

  pilnpi = − xlnx +

  − x ln

  x

  x ln

  x

  

  

  

  −

  

   +

  +

  

  

  +

  ∑

  2

  

   [B.11]

  

   3

    3

    3

    3

  

  i

  If we determine the maximum of S with respect to x, we achieve x = 1/9.

  Try by yourself, just as an exercise.

  Maximizing other functions of pi imposes correlation between the eyes’ color and handedness

  of kangaroos. For example, before the work by Shannon, other information measures had been

  proposed. One of them was −∑ p 2 i. Such function does not give the same result of the Shannon

  i

  Entropy. In fact, maximizing −∑ p 2 i, we obtain x =1 1

  / 2, i.e., a negative correlation between the two

  i

  uncorrelated properties: blue-eyes and left-handedness.

  Appendix B

  511

  Before ending this Appendix, it is important to present another consideration about the “likeli-

  hood” function of equation [B.3]. When the experimental technique, chosen to collect the time-

  resolved signal, taints the data with “normal” noise, the Gaussian distribution is the right function,

  the nonlinear least-squares a good fitting procedure, and the minimization of χ 2 the right way to fit the

  data. However, there are techniques that are better described by the Poisson statistics. An example

  is the Time-Correlated Single Photon Counting technique. In such technique, the time scale is parti-

  tioned in n channels, and the luminescence intensity decay curve is built by collecting the number of

  independent photons emitted in each channel, c , , …, . Such counts follow the Poisson statistics.

  1 c 2

  cn

  In statistics, there exist three probability distributions. The parent distribution is the Binomial one

  (see Figure B.5) representing the number of successes in a sequence of n-independent yes/no experiments, each of which yields the successful outcome with probability p, and the negative result with

  probability q = 1 − p. It is an asymmetric distribution unless p = 1/2. If we repeat the experiment n-times, the average value of the positive outcome is ν = np, and its standard deviation is σ = npq .

  When the number of experiments is enormous, i.e., n → ∞, the Binomial distribution becomes symmet-

  ric. It can be substituted by the Gaussian distribution (see Figure B.5) that is centered at ν = N, which is also the mean value; σ is its standard deviation that defines the width of the distribution. When the number of experiments is huge, i.e., n → ∞, and the probability of the positive result is tiny, i.e., p → 0, the Binomial distribution can be replaced by the Poisson distribution (presented in Figure B.5), wherein μ is its average value of ν , and σ = µ is the standard deviation. The Poisson distribution describes appropriately all those experiments where random events occurring with small probabilities p are considered.

  Examples of such events are radioactive decays and the counting of emitted photons in a broad array of

  channels. Usually, the Poisson distribution is asymmetric, but when µ → ∞, it becomes symmetric, and

  it can be replaced by a Gaussian distribution having the same average and standard deviation.

  In the case of the Single-Photon Counting Technique (Bajzer and Prendergast 1992), the prob-

  ability of the observed counts c , p( c α τ ), follows the Poisson distribution:

  i

  i ; ( )

  e−〈 ci〉 c ci

  p( c

  i

  )

  i ;α =

  〈 〉 [B.12]

  ci !

  where 〈 ci〉 is the expected value of the number of counts in the i- th channel. Since the counts are

  independent, the likelihood will be the joint probability

  Pr( D h) = ∏ p( ci;α) [B.13]

  i

  The value 〈 ci〉 is modeled by the function R h for i = 1,…, n. The best estimates of the R h will i

  i

  be those values that maximize the likelihood and the entropy. Note that the log-likelihood func-

  tion, ln  Pr( D h)

  

  , attains its maximum for the same value of h, as the likelihood function. It is

  Binomial distribution

  P( v, n) =

  n!

  pv (1 − p) n− v

  v!( n − v)!

  Gaussian

  Poisson

  distribution

  distribution

  n→∞

  n→∞

  p→0

  e−( v − N)/2 σ 2

  μ→∞

  μv

  p( v) =

  1

  p( v) = e− μ v!

  σ√2 π

  FIGURE B.5 Formula of the three distributions used in statistics to describe random events.

  512

  Appendix B

  customary to determine the best h by minimizing −ln  Pr( D h), which is equivalent to minimizing the Poisson deviance

  D(α) = 2

  c

  ∑{

  }

  iln  c

   i / Rhi   c

   −  i − Rhi  [B.14]

  i

  The Poisson deviance is close to the Gaussian χ 2 when the counts are very high. This result is in

  agreement with the consideration that a Poisson distribution can be replaced by a Gaussian one

  when the average value in each channel goes to infinity.

  B.1 HINTS FOR FURTHER READING

  Further information regarding the Maximum Entropy Method applied to time-resolved spectros-

  copy, and image processing can be found on the website of Center for Molecular Modeling of the

  National Institutes of Health (https://cmm.cit.nih.gov/maxent/). It is possible to download the software MemExp by P. J. Steinbach, for the analysis of the kinetic data.

  Appendix C: Fourier Transform

  of Waveforms

  Computers are useless. They can only give you answers.

  Pablo Picasso (1881–1973 AD)

  Almost everything in the universe can be described as a waveform. Suffice to think that quantum

  particles are wave packets. A waveform is a curve showing the shape of a wave as a function of time

  or space or some other variables.

  The Fourier Transform (Bracewell 1986) is a powerful mathematical tool that allows describing

  any waveform as a sum of sinusoids, i.e., as a sum of sine and cosine functions.

  Let us remember the general expression and features of a sinusoid, like a sine function:

  y = A sin(2πω t +ϕ) [C.1]

  The function y has a frequency ω that is the inverse of the time it takes to complete one period of

  oscillation. The term A in equation [C.1] is the amplitude of y (see Figure C.1), and φ is its phase. In the bottom graph of Figure C.1, the phase difference (Δ φ) between two sinusoids, having the same frequency and the same amplitude, is highlighted.

  When we encounter a waveform like that depicted in Figure C.2, we ask ourselves: “If we describe

  the signal S as a sum of sinusoids having different frequencies, what frequencies are needed, and

  what is the distribution of the amplitudes and phases as the functions of the frequency?” This ques-

  tion has an answer because Fourier’s Theorem states that any signal of zero mean value can be

  represented as a unique sum of sinusoids.

  When we sample the signal S with a time resolution of ∆ t (see Figure C.2), and we collect N data, the total time window is TW = N * t

  ∆ . Inevitably, the smallest period we can infer from our set of

  data is TMIN = 2 t

  ∆ . In other words, the largest frequency we can measure in our series of sinusoids

  describing the data is ν MAX = 1 2

  / t

  ∆ . If the original signal S contains periods shorter than TMIN = 2 t

  ∆

  or frequencies higher than ν MAX, our measurement is said to be aliased. If we want to evaluate short-

  est periods, we must reduce the time interval ∆ t . The longest period we can obtain unambiguously

  from a windowed trend is one that lasts the total sampling time, that is TMAX = N * t

  ∆ ; in other

  words, the smallest frequency we can infer is ν MIN = 1/ N t

  ∆ . In addition, T defines our resolu-

  MIN

  tion in the determination of the period. Analogously, ν MIN defines the resolution in the frequency

  determination. It derives that the maximum number of the sinusoids in the Fourier Transform will

  be N/2. This statement makes sense because we have N data, and for the definition of each sinusoid

  we need to determine two parameters: its amplitude and its phase. For each frequency ν = n/ N t

  ∆ ,

  with n = 1, 2, …, N/2, its amplitude is

  A

  2

  2

  n =

  an + bn [C.2]

  and its phase is

   b 

  ϕ

  n

  n = −arctan 

   [C.3]

   n

  a 

  where the coefficients, a and b , are given in terms of the original signal S( t) as

  n

  n

  513

  514

  Appendix C

  1.0

  1/ ω

  0.5

  A

  y 0.0

  −0.5

  −1.0

  0.0

  0.5

  1.0

  1.5

  2.0

  1.0

  0.5

  y 0.0

  −0.5

  Δ ϕ

  −1.0

  0.0

  0.5

  1.0

  1.5

  2.0

  t

  FIGURE C.1 Shape and features of the sinusoids.

  4

  Δ t

  2

  0

  S

  −2

  −4

  −6

  Time

  FIGURE C.2 An example of a waveform regarding the signal S as a function of time. Δ t represents the sampling period.

  N

  2

  2π ni

  N

  

  2

  2π ni 

  an =

  

  S( i t

  ∑ ∆ )cos  bn =

  

  S( i t

  ∑ ∆ )sin

  N

  

   [C.4]

   N 

  N

   N 

  i=1

  i=1

  Expressions [C.4], [C.2], and [C.3] are referred to as the Discrete Fourier Transform (DFT).

  The original set of data S( iΔ t) is finally expressed in terms of the sum of the N/2 sinusoids: A

  N /2

  S( i t

  ∆ ) = 0 +

  An cos(2πν ni t

  ∆ +ϕ n) [C.5]

  2

  ∑ n=1

  Appendix C

  515

  Frequency

  400 0.0

  0.1

  0.2

  0.3

  0.4

  0.5

  200

  Phase

  0

  1.2

  1.0

  0.8

  0.6

  Amplitude 0.4

  0.2

  0.0 0.0

  0.1

  0.2

  0.3

  0.4

  0.5

  Frequency

  FIGURE C.3 Discrete Fourier Transform of the data plotted in Figure C.2.

  Equation [C.5] is called the Inverse Discrete Fourier Transform. The term A 0 2

  / represents any mean

  value of the signal which is not accounted for in the other sinusoids. It comes from the definitions of

  a and b (cf. equation [C.4]) when n

  n

  n

  = 0, which corresponds to frequency 0.

  When we decompose our signal S into its component sinusoids, the resulting graph reporting

  amplitude or phase as a function of frequency is referred to as its Fourier spectrum. In Figure C.3, we report the Fourier spectrum of the 53 data plotted in Figure C.2. It consists of 27 frequencies included between 0 and 0.5.

  Usually, the DFT is computed by an algorithm known as the Fast Fourier Transform (FFT). The

  key advantage of the FFT over the DFT is that the operational complexity decreases from O(N2) for

  a DFT to O(Nlog (N)) for the FFT.

  2

  Appendix D: Errors and Uncertainties

  in Laboratory Experiments

  Those sciences are vain and full of errors which are not born from experiment, the mother of

  certainty.

  Leonardo da Vinci (1452–1519 AD)

  Science is said to be exact not because it is based on infinitely exact assertions but because its rigor-

  ous methodology allows estimating the extent of the uncertainty associated with any determination.

  D.1 ERRORS IN DIRECT MEASUREMENTS

  Many scientific assertions are grounded in experiments. Leonardo da Vinci was used to say that

  scientific knowledge without experiments is vain and full of mistakes. However, we are aware that

  even when we perform experiments, we make errors.

  Usually, an experiment involves four protagonists: the observer raising a question and looking

  for an answer; the object or the phenomenon that is inquired; an instrument, and the surrounding

  environment (see Figure D.1).

  The Galilean experimental method requires that the object or the phenomenon under study is ana-

 

Add Fast Bookmark
Load Fast Bookmark
Turn Navi On
Turn Navi On
Turn Navi On
Scroll Up
Turn Navi On
Scroll
Turn Navi On
155