Professional Documents
Culture Documents
Quantum Games Entropy
Quantum Games Entropy
Abstract
This paper introduces Hermite’s polynomials, in the description of quan-
tum games. Hermite’s polynomials are associated to the gaussian probabil-
ity density . The gaussian probability density represents (minimum disper-
sion ) and permit us to analyze the convergence to the Nash’s equilibrium
(minimun entropy ME or maximum utility MU). The ME concept is re-
lated to Quantum Games. Some questions arise after carrying out this
exercise: i) What does the Heisemberg’s uncertainly principle represent in
Game Theory and Time Series?, and ii) What does the postulates of the
Quantum Mechanics indicate in Game Theory and Economics?.
1 INTRODUCTION
The quantum games and the quantum computer are deeply related. Quantum
Computer is one of the modern paradigms in Computer Science and Game Theory
[10], [9], [8], [26], [44] [45], [58]. Quantum Computer will allow to increase the
processing speed and the effective search in the databases. The quantum games
allow to incorporate the algorithms of Quantum Theory to Game Theory [8], this
simple extrapolation allows to solve the Prisoner’s Dilemma and demonstrate that
the cooperative equilibrium is viable and stable with probability different to zero.
1
“At the most abstract level, game theory is about numbers of entities that
are efficiently acting to maximize or minimize[10]. For a quantum physicist, it is
then legitimate to ask: what happens if linear superpositions of these actions are
allowed for?, that is, if games are generalized into the quantum domain. For a
particular case of the Prisoner’s Dilemma, we show that this game ceases to pose
a dilemma if quantum strategies are implemented for.” They show that classical
strategies are particular cases of quantum strategies.
Eisert. Wilkens, and Lewenstein 2001, not only give a physical model of
quantum strategies but also gives the idea of identifying moves with quantum op-
erations and quantum properties. This approach appears to be fruitful in at least
two ways. By the Þrst hand several recently proposed applications of quantum
information theory, can already be conceived as competitive situations, where
several parts with more or less opposed motives to interact. These parts may
apply quantum operations on a bipartite quantum system. On the other hand a
generalization of the theory of decisions in the domain of quantum probabilities
seems interesting, as the roots of game theory are partly in probability theory.
In this context it is of interest to investigate what solutions are attainable if su-
perpositions of strategies are allowed. A game is also related to transference of
information. It is possible to ask, what happens if these carriers of information
are taken to be quantum systems?, quantum information being a fundamental
notion of information. The Nash’s equilibrium concept in quantum games is es-
sentially the same that in game theory but one difference is that the strategies
are in function of quantum properties of the physical system.
This paper essentially explicits the existent relationship among Quantum Me-
chanics, Nash’s equilibria, Heisenberg Uncertainty Principle, Minimum Entropy
and the Time Series. The Heisenberg’s uncertainty principle is one of the angular
stones in Quantum Mechanics. One application of the uncertainty principle in
Time Series is related to Spectral Analysis “The more precisely the aleatory
variable VALUE is determined, the less precisely the frequency VALUE is known
in this instant, and conversely”. This principle indicates that the product of the
standard deviation of one aleatory variable xt by the standard deviation of its
frequency w is greater or equal to 1/2.
This paper is organized as follows:
Section 2: Quantum Games and Hermite’s Polynomials, Section 3: Time Se-
ries and Heisenberg’s Principle, Section 4: Quantum Games Application, Section
5: Application of Heisenberg’s Principle in Time Series, Section 6: Conclusion
and Section 7: Appendix
2
where k ∈ K, skji ∈ Sk , jk = 1, ..., lk and S = ΠK Sk set of pure strategy proÞles
with s ∈ S an element of that set, l = l1 , l2 , ..., ln represent the cardinality of S
. The vectorial function v : S → Rn associates every proÞle s ∈ S the vector of
utilities v(s) = (v 1 (s), ..., vn (s))T , where v k (s) designates the utility of the player
k facing the proÞle s. In order to get facility of calculus we write the function
v k (s) in one explicit way vk (s) = v k (j1 , j2 , ..., jn ).The matrix vn,l represents all
points of the cartesian product Πk∈K Sk . The vector vk (s) is the k− column of v.
If the mixed strategies are allowed then we have:
( lk
)
X
∆(Sk ) = pk ∈ Rlk : pkjk = 1
jk =1
the unit simplex of the mixed strategies of player k ∈ K, and pk = (pkjk ) the
probability vector. The set of proÞles in mixed strategies is the polyhedron ∆
with ∆ = Πk∈K ∆(Sk ), where p = (p1j1 , p2j2 ..., pnjn ), and pk = (pk1 , pk2 , ..., pkln )T .
Using the Kronecker product ⊗ it is possible to write1 :
lk
X
uk = ukjk (p(−k) , v(s))pkji
jk =1
0
u = vp
¡ ¢
uk = lk ⊗ vk p(−k)
The triplet (K, ∆, u(p)) designates the extension of the game Γ to the mixed
strategies. We get Nash’s equilibrium
¡ ¢∗ (maximization of utility) if and only if, ∀k,
p, the inequality uk (p∗ ) ≥ uk ( pk , p(−k) ) is respected .
Another way to calculate the Nash’s equilibrium [52], is equaling the values
of the prospective utilities of each strategy when it is possible.
1
We use bolt in order to represent vector or matrix.
3
¡ ¢ ¡ ¢ ¡ ¢
uk1 p(−k) , v(s) = uk2 p(−k) , v(s) = ... = ukjk p(−k) , v(s)
lk
X
pkjk = 1 ∀k = 1, ..., n
jk =1
lk ³
X ´2
¡ ¢
σ 2k = ukjk p(−k) , v(s) − uk pkjk = 0
jk =1
¡ ¢∗
If the resulting system of equations doesn’t have solution p(−k) then P we
propose the Minimum Entropy Method. This method is expressed as M inp ( k Hk (p)) ,
where σ 2k (p∗ ) standard deviation and Hk (p∗ ) entropy of each player k.
¢∗ ³¡´
σ 2k (p∗ ) ≤ σ 2k
pk , p(−k) or
³¡ ¢∗ ´
∗ k (−k)
Hk (p ) ≤ Hk p , p
Z +∞
Hk = − p(x) ln(p(x))dx
−∞
Z +∞ Ã Ã ! µ ¶2 !
1 x − uk
Hk = − p(x) ln p − √ dx
−∞ 2πσ 2k 2σ k
4
developing the integral we have
³√ ´ Z +∞
Hk = ln 2πσ k p(x)dx
−∞
Z +∞
1
+ 2 (x − uk )2 p(x)dx
2σ k −∞
³√ ´ µ ¶
σ 2k 1 √
= ln 2πσ k + 2 = + ln 2π + ln σ k
2σ k 2
For a game to n− players the total entropy can be written as continues:
X
n µ¶
1 √
Hk = n + ln 2π + ln(Πnk=1 σ k )
2
k=1
à !
X
M inp Hk (p) ⇔ Minp (Πk σ k (p))
k
k −k ) k −k )
eλujk (p eλujk (p
pkjk = Pl λukj (p−k )
=
k
e k Zk
jk =1
³ ´ P k −k
where Zk λ, uk (p−k ) = ljkk =1 eλujk (p ) represents the partition utility func-
tion.
The entropy Hk (pk ), expected utility E{ukjk (p−k )} = uk (p) , and variance
V ar{ukjk (p−k )} will be different for each player k.
5
lk
X
Hk (pk ) = − pkjk ln(pkjk )
jk =1
lk
X
uk (p) = pkjk ukjk
jk =1
lk
X
© k −k ª
V ar ujk (p ) = pkjk (ukjk − uk )2
jk =1
Using the explicit form of pkjk , we can get the entropy, the expected utility
and the variance:
6
Remark 1 The entropy Hk for gaussian probability density and multinomial
¡ √ ¢ P k
logit is written as 12 + ln 2π + ln σ k and ln( ljkk =1 eλujk ) − λuk
Case 3 The special case of Minimum Entropy is when σ 2k = 0 and the utility
function value of each strategy ukjk (p(−k) , v(s)) = uk , is the same ∀jk , ∀k.
uk1 (p(−k) , v(s)) 6= uk2 (p(−k) , v(s)) 6= ... 6= ukn (p(−k) , v(s))
X lk
pkjk = 1
jk =1
lk
X
σ 2k = (ukj (p(−k) , v(s)) − uk )2 pkjk > 0
jk =1
2
Theorem 2 Minimum Dispersion. The gaussian density function ³ f (x) = ´ |φ(x)|
represents the solution of one differential equation given by 2σx2 + ∂x
∂
φ(x) =
x
0 related to minimum dispersion of a¡ lineal¢ combination between the stochastic
∂
variable x, and a Hermitian operator −i ∂x .
bB
Proof. This proof is extended in [51], [7]. Let A, b be Hermitian operators
2
which do not commute, because we can write,
h i ³ ´
b b b b b
A, B = AB − B A = iCb
2 eb
b have the next property: A eb
b∗ , the transpose operator A
The hermitian operator A =A is
b∗
equal to complex conjugate operator A .
7
Easily we can veriÞed that the standard deviations satisfy the same rule of
commutation:
h i h i h i
b b b b b b
A, B = 4A, 4B = A − A, B − B
h i
b 4B
4A, b = 4A4
b Bb − 4B4
b A b = iC
b
the expression J(α) has the minimum value when J 0 (α) = 0, thus the value
of α and Jmin are
D E D E
b b
i 4A4B − 4B4A b b Cb
α = − ¿³ ´2 À = ¿³ ´2 À
2 4Ab 2 4A b
D E D E
b2
C ¿³ ´2 À Cb2
Jmin = ¿³ ´2 À + b
4B − ¿³ ´2 À ≥ 0
4 4A b 2 4A b
simplifying Jmin
¿³ ´2 À ¿³ ´2 À
b b 1 D b2 E
4A 4B ≥ C
4
1
σ Abσ Bb ≥
2
We are interested in knowing the explicit form of the function φ(x)
£ when ¤the
b b ∂ ∂
inequality is transformed into equality and A = x, B = −i ∂x , where x, −i ∂x =
i for which we need to solve the following differential equation.
8
³ ´
b + i4B
α4A b φ(x) = 0
D E
Cb
¿³ b + i4B
´2 À 4A b φ(x) = 0
2 4A b
D E D E
b b ∂ b
if A = x, B = −i ∂x , and B = 0 then C b = 1 and
µ ¶
x − hxi ∂
+ φ(x) = 0
2σ 2x ∂x
the solution is
µ ¶1/2
1 (x−hxi)2
φ(x) = √ e− 4σ2
σ x 2π
(x−hxi)2
1 −
f (x) = |φ(x)|2 = √ e 2σ2x
σ x 2π
Remark 2 The gaussian density function f(x) is optimal in the sense of mini-
mum dispersion.
According to the last theorem ukjk = x is a gaussian aleatory variable with
h i
probability density fk (x), ∀k, and for case of calculation E ukjk = uk . We can
write
2
(x−uk )
1 −
2σ 2
fk (x) = √ e k
σ k 2π
Corollary 1 The central limit theorem which establishes that “The mean uk of
random variable ukjk has normal density function”, is respected in Minimum
Entropy Method.
9
Table 1: Quantum Mechanic and Game Theory properties
10
Definition 1 (Tactics of Sub- Strategies) For a correct description of the strategy
evolution jk , it is necessary the introduction of sub strategies mkjk ∈ Mjkk , ∀k, jk
¯ ¯
¯ k ¯2
with a probability given by ¯¯bmk ¯¯ where mkjk ∈ M k . Let Mjkk be the set of sub-
jk
strategies with ∪ljkk =1 Mjkk = M k , M k = {0, 1, ..., ∞} and Mjkk ∩Mjkl = ∅ when k 6=
l. A strategy jk is constituted of several sub strategies mkjk . With the introduction
of sub strategies the¯ probability ¯ of each strategy is described in the following way
¯ k ¯2 P ¯ k ¯2
¯bj ¯ = ¯ ¯
k mkj ∈Mjk ¯bmkj ¯ and we don’t need any approximation of the function
P∞ k k k
P∞
mkj =0 cmkj (t)ϕmkj (x, λ), where ψ(x, λ) = mk =0 cmk (t)ϕmkj (x, λ), cmkj (t) =
k
¯ k ¯2 ¯k ¯2 ¯ jk ¯2 jk k k
¯ ¯ ¯ ¯ P ¯ ¯
bkjk eiwk t ,¯¯c2mk ¯¯ = ¯¯bkmk ¯¯ = pkmk and ∞ ¯ k ¯
mk =0 ¯bmk ¯ = 1, ∀k.
jk jk jk jk jk
Theorem 3 The normal probability density ρ(u) = |ψ (x, λ)|2 ¯can be obtained
¯2
using Hermite orthogonal polynomials Hk (x). The parameter ¯bkjk ¯ , indicates
the probability value of playing jk strategy. The state function ϕk (x, λ) measure
the behavior of jk strategy (jk one level in Quantum Mechanics).
P∞ The dynamic
behavior of jk strategy can be written as ψ(x, λ) = mkj =0 cmkjk (t)ϕmkj (x, λ).
P k k
One approximation is ψ(x, λ) = mk ∈Z + cmkj (t)ϕmkj (x, λ)
jk k k
1 −( √x −λk )2
ψ(x, λ) = p √ e 2
σ k 2π
∞
X mk µ ¶
1 (λk ) jk − x2 x
= p √ e 2 Hmkj √
k σ k 2π mkjk ! k 2
m =0jk
√u , µ
where: x = 2σk
λ= 2σk
, and
D E
ϕmkj , ϕmkj = 1,
D k k
E
ϕmkj , ϕmkj = 0, ∀mkjk 6= mkjl
k l
2
³ ´
− x2
e Hmkj √x2
ϕmkj (x, λ) = q √
k
k mk √
mkjk ! 2 jk π
11
1 x 2
p √ e−( 2 −λk )
√
(Q1)
σ k 2π
à k
! µ ¶
∞
X (λk )mjk 2 x
− x2
= p √ e Hmkj √
mk =0
mk
jk ! σ k 2π k 2
jk
1 −
(x−λk )2
p √ e 2 (Q3)
σ k 2π
q ³ ´
k √ mkj −1 e− x22 H k √x
X∞ m
(λk )q 2 k
jk mj 2
= q k
k √ √ m k √
mkj =0 mjk ! σ n mkj ! 2 jk π
k k
q
√ mkj −1
mkj
(λk ) k 2 k
bmkj = q √ (Q4)
k
mkjk ! σ k
x2
³ ´
e− 2 Hmkj √x2
ϕmkj (x, λk ) = q √
k
(Q5)
k mk √
mkjk ! 2 jk π
¯ ¯2
¯ ¯
The ¯bmkj ¯ parameter represents the probability of mkjk sub strategy of each
k
k player. The utility of vNM can be written as:
12
h i lk
X ¯ k ¯2 k
k k −k
E ujk = u (p ) = ¯bj ¯ uj (Q5.1)
k k
jk =1
¯ k ¯2 X ¯ ¯2
¯b ¯ = ¯ k ¯
jk ¯bmkj ¯ (Q5.2)
k
mkj ∈Mjk
k k
∂
(x + ∂x )
aϕmkj (x, λk ) = ¡√ ¢ 3/2
ϕmkj (x, λk ) (Q6)
k k
2
√
aϕmkj (x, λk ) = nϕmkj −1 (x, λk ) (Q6.1)
k k
∂
(x+ ∂x )
The equation (Q6.1) indicates that the operator a = √ 3/2 decrease de level
( )
2
of k player. The operator a can decrease the sub strategy description mkjk →
mkjk − 1.
∂
¡¢
+ x − ∂x
a ϕmkj (x, λ) = ¡√ ¢3/2 ϕmkj (x, λ) (Q7)
k k
2
√
a+ ϕmkj (x, λ) = n + 1ϕmkj +1 (x, λ) (Q7.1)
k k
∂
(x− ∂x )
The equation (Q7.1) indicates that the operator a+ = √ 3/2 increase the
( 2)
level of k player. The operator a+ can increase the sub strategy description
mkjk → mkjk + 1.
The Hermite polynomials form an orthogonal base and they possess desir-
able properties that permit to express a normal probability density as a linear
combination of them.
13
mkj 2
x2 d
k e−x
Hmkj (x) = (−1)e k (PH1)
k
dxmjk
k mkj (mkjk − 1) k
Hmkj (x) = (2x)mjk − k (2x)mjk −2
k 1!
mkjk (mkjk − 1)(mkjk − 2)(mkjk − 3) k
+ (2x)mjk −4 +
2!
Taking the explicit value of mkjk −derivative of (PH1):
mkj
d k Hmkj (x) k
mkj
k
= mkjk !2mjk (PH2)
dx k
mk
d jk Hmkj ( √x2 ) √ mk
k
mkj
= mkjk ! 2 jk (PH2.1)
dx k
Decreasing operator:
dHmkj (x)
k
= 2nHmkj −1 (x) (PH3)
dx k
Increasing operator:
d
2x(x − )H k (x) = Hmkj +1 (x) (PH4)
dx mjk k
Recurrence formula:
Normalization condition:
Z +∞
2
e−x Hm
2
k (x)dx (PH6)
jk
−∞
Z +∞
k
dmjk Hmkj (x) k √
−x2
= e mkj
k
dx = mkjk !2mjk π
−∞ dx k
The next condition expresses the orthogonal property of Hermite’s polyno-
mials:
D x2 2 E
− 2 − x2
e Hmkj (x), e Hmkj (x)
k k
Z +∞
2
= e−x Hmkj (x)Hmkj (x)dx = 0 ∀mkjk 6= mkjl
k l
−∞
14
Generating Function is obtained using the Taylor development.
X (λk )mkjk
x2 −(x−λk )2
F (x, λk ) = e e = Hmkj (x) (PH7)
mkjk ! k
mkj
k
Let n be players with k = 1, ..., n , lk strategies and mkjk ∈ Mjkk of each player
k.
Superposition Principle:
∞
X D E
k
ψ = ckmk ϕmkj , ϕmkj , ϕmkj = δ mkj mkj (QG1)
j k k k l k l
mkj =0
k
Time Evolution:
√
ckmk = bkmk eiwk t , i= −1 (QG2)
jk jk
Inner Product:
15
D E
ckmk = ψ k , ϕmkj (QG3)
jk k
Quantum Property
¯ ¯ ¯ ¯2
¯ k ¯2 ¯ ¯
¯cmk ¯ = ¯bkmk ¯ = pkmk (QG4)
jk j j k k
∞
X lk
X
pkmk = pkj = 1
j k k
mkj =0 jk =1
k
X
pkj = pkmk
k jk
mkj ∈Mjk
k k
Mean Value
®
E [xk ] = xk = ψ k |x| ψ k (QG5)
Normality Condition
® ¯ ¯2
ρk = ψ k , ψ k = ¯ψ k ¯ (QG6)
• A set Ω that represents the set of all possible outcomes of a certain random
experiment.
— ∅∈=
— A ∈ = =⇒ Ac ∈ =
— A1 , A2 , ... ∈ = =⇒ ∪∞
i=1 Ai ∈ =
— P (∅) = 0, P (Ω) = 1
16
— If A1 , A2 , ... ∈ = form a Þnite or countably inÞnite collection of dis-
joints sets (that is, Ai ∩ Aj = ∅ if i 6= j) then
∞
X
P (∪∞
i=1 Ai ) = P (Ai )
i=1
The elements of the σ-algebra = are called events and P is called a probability
measure.
X : Ω→R
ω → X(ω)
17
The variable X(t, •) is =−measurable in ω ∈ Ω for each t ∈ T ; henceforth
we shall often follow established convention and write Xt for X(t). When T is
a countable set, the stochastic process is really a sequence of random variables
Xt1 , Xt2 , ..., Xtn , ..
Z +c
1 e−iwa − e−iwb
F (a, b] = lim h(w)dw
c→∞ 2π −c iu
for all points a, b (a < b) at which F is continuous. If in addition, h is Lebesgue
integrable on (−∞, ∞) , then the function f given by
Z +∞
1
f (x) = e−iwx h(w)dw
2π −∞
R +x
is a density of F, that is, f is nonnegative and F (x) = −∞ f (t)dt for all
x; furthermore, F 0 = f everywhere. Thus in this case, f and h are “Fourier
transform pairs”:
Z +∞
1
h(w) = z(f (x)) = eiwx f(x)dx
2π −∞
Z +∞
1
f(x) = z−1 (h(w)) = e−iwx h(w)dw
2π −∞
18
If the functions ψ(x = ±∞) = 0 and φ(w = ±∞) = 0 then f (x = ±∞) = 0
and h(w = ±∞) = 0.
Proof. The next proof is based in [13].
Let the function ψ(x) deÞned as,
Z +∞
−1 1
ψ(x) = z [φ(w)] = √ φ(w)e−iwx dw ∀x, ψ(x = ±∞) = 0 (Q0)
2π −∞
Z +∞
1
φ(w) = z [ψ(x)] = √ ψ(x)eiwx dx ∀x, φ(w = ±∞) = 0 (Q1)
2π −∞
Z +∞
∂ n φ(w) in
n
= √ xn ψ(x)eiwx dx = in z [xn ψ(x)] (Q2)
∂w 2π −∞
Z
n
∂ ψ(x) (−i)n +∞ n
n
= √ w φ(w)e−iwx dw = (−i)n −1 z [wn φ(w)]
∂x 2π −∞
where:
µ ¶ Z +∞ n
∂ n ψ(x) 1 ∂ ψ(x) iwx
z = √ e dx = (iw)n φ(w) (Q3)
∂xn 2π −∞ ∂x n
µ n ¶ Z +∞ n
∂ φ(w) 1 ∂ φ(w) −iwx
z−1
n
= √ n
e dw = (−ix)n ψ(x)
∂x 2π −∞ ∂x
By Parseval Plancherel theorem
Z +∞ Z +∞
∗
φ(w)φ(w) dw = ψ(x)ψ(x)∗ dw (Q4)
−∞ −∞
19
Z +∞
φ(w)φ(w)∗ dw = 1 (Q6)
−∞
Z +∞
ψ(x)ψ(x)∗ dw = 1
−∞
Let σ 2w
be the standard deviation of frequency w, and σ 2x the standard devi-
ation of the aleatory variable x.
Z +∞
σ 2w = w 2 φ(w)φ(w)∗ dw (Q7.1)
−∞
Z +∞
σ 2x = x2 ψ(x)ψ(x)∗ dx (Q7.2)
−∞
with Parseval Plancherel theorem
Z +∞ Z +∞
∂ψ(x) ∂ψ(x)∗
dx = (wφ(w)) (wφ(w))∗ dw (Q8.1)
−∞ ∂x ∂x −∞
Z +∞ Z +∞
ψ 0 ψ 0∗ dx = w2 φ(w)φ(w)∗ dw (Q8.2)
−∞ −∞
f 0 = ψ 0 ψ ∗ + ψψ 0∗
Z +∞ Z +∞
∗ 0 ∗ 0∗
A+A = xdx (ψ ψ + ψψ ) = xf 0 (x)dx
−∞ −∞
Z +∞
A + A∗ = − f (x)dx = −1 (Q11)
−∞
20
Notice that A + A∗ = −1 and using the complex number property 2 Re (A) =
−1 or 2 |A| cos θ = −1 where θ phase of A, we can write 4 |A|2 = cos12 θ . Since
−1 ≤ cos θ ≤ 1 then |A|2 ≥ 14 and therefore:
1
σ 2x σ 2w ≥ (Q12)
4
x 2 2 2
w σ
Example 1 If f (x) = √1 e− 2σ2 then g(w) = √σ e− 2
σ 2π 2π
M
X
xt = µ + {aj cos(wj (t − 1)) + bj sin(wj (t − 1))} + ut or (Q14)
j=1
xt = c + φ1 xt−1 + ... + φp xt−p + εt + θ1 εt−1 + ... + θq εt−q
The only condition that imposes the Possibility Theorem is exhaustive and
exclusive σ xσ w ≥ 12 , because it includes the Minimum Entropy Theorem or
equilibrium state theorem σ x σ w = 12 . It is necessary to analyze the different
cases of the time series xt using the Possibility Theorem. A time series xt evolves
in the dynamic equilibrium if and only if σ x σ w = 12 . A time series evolves out of
the dynamic equilibrium if and only if σ x σ w > 12 .
21
Table 2: Evolution of Time Series, out of the equilibrium σ x σ w > 12 .
Z +∞
H(W ) = dwp(w) log(p(w)) (Q15)
−∞
Z +∞
H(X) = dxp(x) log(p(x))
−∞
µ ¶
1
H(W ) + H(X) ≥ 1 + log
2
Example:
When ψ(x) and φ(w) are Gaussian H(W ) = B + Log(σ W ) and H(X) =
C + Log(σ X ),Hirschman’s inequality become Heisenberg, then inequalities are
transformed in equalities and the minimum uncertainly is minimum entropy. “in
Quantum Mechanic minimum uncertainty product” also obeys a minimum en-
tropy sum.
22
4 Applications of the Models
4.1 Hermites’s Polynomials Application
Let Γ = (K, S, v) be a game to 3−player, with K the set of players k = 1, 2, 3.
The Þnite set Sk of cardinality lk ∈ N is the set of pure strategies of each player
where k ∈ K, skjk ∈ Sk , jk = 1, 2, 3 and S = Πk Sk set of pure strategy proÞles
with s ∈ S an element of that set, l = 3 ∗ 3 ∗ 3 = 27 represent the cardinality of S
. The vectorial function v : S → R3 associates every proÞle s ∈ S the vector of
utilities v(s) = (v1 (s), ..., v 3 (s)), where v k (s) designates the utility of the player
k facing the proÞle s. In order to get facility of calculus we write the function
v k (s) in one explicit way v k (s) = v k (j1 , j2 , ..., jn ).The matrix v3,27 represents all
points of the Cartesian product ΠK Sk . The vector vk (s) is the k- column of v.
The graphic representation of the 3-players game (Figure 1).
23
Table 4: Maximum Utility maxp (u1 + u2 + u3 )
1 2 3
Max u +u +u = 20,369
1 2 3
u u u
8,7405 9,9284 1,6998
σ1 σ2 σ3
6,3509 6,2767 3,8522
H1 H2 H3
4,2021 4,1905 3,7121
Kroneker Products
u (j1,j2,j3)
u2(j1,j2,j3)
u3(j1,j2,j3)
v1(j1,j2,j3)
v2(j1,j2,j3)
v3(j1,j2,j3)
24
Table 5: Minimum Entropy minp (σ 1 σ 2 σ 3 ) ⇔ minp (H1 + H2 + H3 )
1 2 3
u u u
1,917 1,7668 4,984
σ1 σ2 σ3
2Ε−07 0,45383 2,111
H1 H2 H3
0,035 1,74694 3,131
u2(j1,j2,j3)
u3(j1,j2,j3)
v1(j1,j2,j3)
v2(j1,j2,j3)
v3(j1,j2,j3)
25
Table 6: Minimum Entropy: Stone-paper-scissor minp (H1 + H2 + H3 )
1 2 3
u u u
0,5500 0,5500 0,5500
1 2 3
σ σ σ
0,0000 0,0000 0,0000
H1 H2 H3
0,0353 0,0353 0,0353
Kroneker Products
u (j1,j2,j3)
u2(j1,j2,j3)
u3(j1,j2,j3)
v1(j1,j2,j3)
v2(j1,j2,j3)
v3(j1,j2,j3)
1 2 3 1 2 1 3 2 3
j1 j2 j3 p p p p j1p p j1p p j2p
1
j1 j2 j3 j2 j3 j3
stone stone stone 0 0 0 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,0000 0,0000 0,0000
stone stone paper 0 0 1 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,0000 0,0000 0,1111
stone stone sccisor 1 1 0 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,1100 0,1100 0,0000
stone paper stone 0 1 0 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,0000 0,1100 0,0000
stone paper paper 0 1 1 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,0000 0,1100 0,1111
stone paper sccisor 1 1 1 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,1100 0,1100 0,1111
stone sccisor stone 1 0 1 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,1100 0,0000 0,1111
stone sccisor paper 1 1 1 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,1100 0,1100 0,1111
stone sccisor sccisor 1 0 0 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,1100 0,0000 0,0000
paper stone stone 1 0 0 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,1100 0,0000 0,0000
paper stone paper 1 0 1 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,1100 0,0000 0,1111
paper stone sccisor 1 1 1 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,1100 0,1100 0,1111
paper paper stone 1 1 0 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,1100 0,1100 0,0000
paper paper paper 0 0 0 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,0000 0,0000 0,0000
paper paper sccisor 0 0 1 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,0000 0,0000 0,1111
paper sccisor stone 1 1 1 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,1100 0,1100 0,1111
paper sccisor paper 0 1 0 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,0000 0,1100 0,0000
paper sccisor sccisor 0 1 1 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,0000 0,1100 0,1111
sccisor stone stone 0 1 1 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,0000 0,1100 0,1111
sccisor stone paper 1 1 1 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,1100 0,1100 0,1111
sccisor stone sccisor 0 1 0 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,0000 0,1100 0,0000
sccisor paper stone 1 1 1 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,1100 0,1100 0,1111
sccisor paper paper 1 0 0 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,1100 0,0000 0,0000
sccisor paper sccisor 1 0 1 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,1100 0,0000 0,1111
sccisor sccisor stone 0 0 1 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,0000 0,0000 0,1111
sccisor sccisor paper 1 1 0 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,1100 0,1100 0,0000
sccisor sccisor sccisor 0 0 0 0,3333 0,3333 0,3300 0,1111 0,1100 0,1100 0,0000 0,0000 0,0000
26
4.2 Time Series Application
The relationship between Time Series and Game Theory appears when we apply
the entropy minimization theorem (EMT). This theorem (EMT) is a way to ana-
lyze the Nash-Hayeck equilibrium in mixed strategies. To introduce the elements
the rationality and equilibrium in the domain of time series can be of a big help,
because it will allow us to study the human behavior reßected and registered
in historical data. The main contributions of this complementary focus to time
series has relationship with econophysics and rationality.
• The human behavior, evolves and it is the result of learning. The objective
of learning is the stability and the optimal equilibrium. Due to the above-
mentioned, we can affirm that if learning is optimal then the convergence
to equilibrium is to a bigger speed that when the learning is sub optimal.
To introduce elements of the Nash’s equilibrium in time series, will allow to
evaluate the learning and the convergence to equilibrium through the study
of historical data.
27
If the γ j ’s represents autocovariances of a covariance-stationary process and
using the Possibility Theorem, then
1 X
T
γbj = (xt − x) (xt−j − x) for j = 0, 1, ..T − 1
T t=j+1
1X
T
x = xt (Q21)
T t=1
28
2X
T
b
aj = xt cos(wj (t − 1)) j = 1, 2, ..., M
T t=1
2X
T
bbj = xt sin(wj (t − 1)) j = 1, 2, ..., M (Q24)
T t=1
1X
T
1 X 2 b2
M
2
(xt − x) = (b
a + bj ) (Q25)
T t=1 2 j=1 j
Example 2 As Þrst step we build four time series knowing all their parameters
aj , bj and wj = ( 2π
T
)j, j = 1, ..., M
Second step, we Þnd the value of the parameters γbj , j = 1, ..., (T − 1) for the
time series xt according to Q21 (Table 7, Figure 6).
Third step, we Þnd Sbx (wj ) (Table 8, Figure 7), only for the frequencies
w1 , w3 , w5 , w12 :
Fourth step, we compute the values σ x , σ w and σ w min = 2σ1x . resp. for y, z, u,
(Table 9).
29
1X
T
σx = (xt − x)2
T t=1
1X
4
σw = (wj − w)2
4 j=1
1
Fifth step, we verify that possibility theorem is respected, σ x σ w ≥ 2
resp for
y, z, u, (Table 9).
Example 3 Let time series vt
1 3,5566 0,2097 0,1698 3,5055 3,1050 2,3498 -1,6010 -1,0000 2,0000 0,0000 3 1,0000 1,0000 0,0000 0,0000
2 -0,0304 -0,0018 0,3396 -0,0287 -0,0160 0,0039 0,0181 4,4253 5,5430 3,5718 3 0,9856 0,8730 0,7507 0,8930
3 8,5707 0,5054 0,5094 7,4824 0,3641 -7,1006 8,4472 -0,9191 -1,3299 -3,2157 3 0,9429 0,5243 0,9919 -0,8039
4 -5,0762 -0,2993 0,6792 -3,9495 2,2851 4,9125 1,4853 2,0619 1,0693 -0,6767 3 0,8730 0,0425 0,5600 -0,1692
5 -9,1882 -0,5418 0,8491 -6,0706 7,6121 4,1376 6,6354 6,2277 5,3810 3,8249 3 0,7780 -0,4502 -0,2519 0,9562
6 4,1666 0,2457 1,0189 2,1847 -4,1515 1,5529 3,9281 -0,7458 -1,4455 -2,7669 3 0,6607 -0,8285 -0,8929 -0,6917
7 -4,0303 -0,2377 1,1887 -1,5028 3,6727 -3,7998 0,5102 0,8480 -0,2853 -1,3339 3 0,5243 -0,9964 -0,9279 -0,3335
8 -8,3009 -0,4895 1,3585 -1,7492 4,9368 -7,2476 6,8796 6,7808 4,7136 3,9678 3 0,3729 -0,9113 -0,3333 0,9920
9 5,6071 0,3307 1,5283 0,2382 -0,7130 1,1825 4,8936 0,9424 -1,8169 -2,2383 3 0,2107 -0,5947 0,4875 -0,5596
10 -1,3549 -0,0799 1,6981 0,1720 -0,5050 0,8054 -0,0583 0,4688 -1,8677 -1,9527 3 0,0425 -0,1272 0,9775 -0,4882
11 -7,6012 -0,4482 1,8679 2,2254 -5,9132 7,5737 6,9285 4,2325 3,7424 3,9963 3 -0,1270 0,3727 0,8041 0,9991
12 7,3787 0,4351 2,0377 -3,3215 7,2723 -5,3286 5,7376 -4,3944 -2,2307 -1,6452 3 -0,2928 0,7779 0,0851 -0,4113
13 3,5415 0,2088 2,2075 -2,1057 3,3394 0,1492 0,7488 -7,7556 -3,4154 -2,5151 3 -0,4502 0,9856 -0,6917 -0,6288
14 -4,7306 -0,2790 2,3774 3,4151 -3,1261 -3,6795 4,5790 -2,1067 2,7204 3,9096 3 -0,5946 0,9429 -0,9991 0,9774
15 7,3117 0,4312 2,5472 -6,0575 1,5420 7,2070 4,8257 -5,6880 -2,4485 -1,0047 3 -0,7219 0,6608 -0,6285 -0,2512
16 0,9479 0,0559 2,7170 -0,8637 -0,2773 0,4973 0,3543 -4,9574 -4,6620 -3,0051 3 -0,8285 0,2109 0,1686 -0,7513
17 -12,9805 -0,7655 2,8868 12,5614 9,3691 3,7957 12,9349 4,4681 1,8877 3,7101 3 -0,9112 -0,2926 0,8513 0,9275
18 -3,0598 -0,1804 3,0566 3,0488 2,9609 2,7876 -1,6015 1,8073 -2,2706 -0,3352 3 -0,9677 -0,7218 0,9563 -0,0838
19 -4,8565 -0,2864 - 4,8390 4,7001 4,4263 -2,5510 -1,6735 -5,4012 -3,4084 3 -0,9964 -0,9677 0,4123 -0,8521
20 -14,5155 -0,8560 - 14,0475 10,4826 4,2574 14,4617 3,4914 1,4109 3,4037 3 -0,9964 -0,9678 -0,4115 0,8509
21 0,7591 0,0448 - -0,6918 -0,2225 0,3977 0,2822 -1,3371 -1,5915 0,3440 3 -0,9678 -0,7222 -0,9560 0,0860
22 5,3938 0,3181 - -4,4692 1,1346 5,3158 3,5689 -6,3602 -5,5360 -3,7134 3 -0,9113 -0,2931 -0,8518 -0,9284
23 -4,5998 -0,2712 - 3,3212 -3,0378 -3,5805 4,4498 0,3719 1,3420 2,9992 3 -0,8286 0,2104 -0,1695 0,7498
24 6,9863 0,4120 - -4,1550 6,5864 0,3007 1,4620 -1,1565 -0,4308 1,0133 3 -0,7220 0,6604 0,6278 0,2533
25 8,7250 0,5145 - -3,9290 8,6000 -6,2953 6,7966 -5,9310 -5,1009 -3,9114 3 -0,5947 0,9428 0,9991 -0,9779
26 -5,6106 -0,3309 - 1,6436 -4,3666 5,5907 5,1089 0,0353 1,6076 2,5082 3 -0,4503 0,9857 0,6924 0,6271
27 3,3472 0,1974 - -0,4256 1,2493 -1,9922 0,1366 -1,4358 1,0674 1,6533 3 -0,2929 0,7783 -0,0842 0,4133
28 6,0537 0,3570 - 0,2561 -0,7664 1,2712 5,2899 -6,9779 -4,2510 -3,9967 3 -0,1272 0,3732 -0,8036 -0,9992
29 -8,4908 -0,5007 - -1,7876 5,0459 -7,4096 7,0264 0,4539 2,0295 1,9449 3 0,0423 -0,1266 -0,9777 0,4862
30 0,7532 0,0444 - 0,2807 -0,6862 0,7103 -0,0970 3,4729 2,6667 2,2457 3 0,2105 -0,5943 -0,4883 0,5614
31 4,3535 0,2567 - 2,2820 -4,3380 1,6264 4,1075 0,1767 -3,2212 -3,9667 3 0,3727 -0,9110 0,3324 -0,9917
32 -10,8746 -0,6413 - -7,1833 9,0127 4,8880 7,8365 7,2184 2,3739 1,3255 3 0,5242 -0,9964 0,9276 0,3314
33 -0,9354 -0,0552 - -0,7277 0,4215 0,9050 0,2757 8,3673 4,0944 2,7733 3 0,6606 -0,8288 0,8933 0,6933
34 4,4848 0,2645 - 3,9149 0,1881 -3,7179 4,4218 -0,4090 -2,2665 -3,8223 3 0,7779 -0,4506 0,2528 -0,9556
35 -9,8590 -0,5814 - -9,2952 -5,1647 1,2608 5,8458 1,1695 2,4138 0,6679 3 0,8729 0,0419 -0,5593 0,1670
36 0,0000 - - - - - - 1,5515 5,1066 3,2210 3 0,9428 0,5239 -0,9918 0,8052
37 - - - - - - - -5,7174 -1,5966 -3,5678 3 0,9856 0,8728 -0,7513 -0,8919
Variance: sx2
E[(xt-E[xt])2] 16,9578
2 2 2 2
(a1 +a3 +b5 +b12 )/2 16,5000
γ0 16,9578
T 37
30
31
Table 9: Verification of Possibily Theorem for series xt , yt , zt , ut
Heisenberg's Uncertainty Principle
2 2
E[x t] 0,0002 sx=E[(x t-E[xt]) ] 4,1180 E[yt] 0,0001 sy=E[(yt-E[yt]) ] 3,2059
2 2
E[w] 0,8915 sw=E[(w-E[w]) ] 0,8129 E[w] 1,1038 sw=E[(w-E[w]) ] 1,3208
2 2
E[zt] 0,0001 sz=E[(zt-E[zt]) ] 2,8674 E[ut] 3,0000 su=E[(ut-E[ut]) ] 0,0000
E[w12] 2,0377 sw=E[(w-E[w]) ]
2
− E[w] ∞ sw=E[(w-E[w]) ]
2
−
yt Vs t
8,0000
6,0000
4,0000
2,0000
dd
yt
0,0000
-2,0000 0 5 10 15 20 25 30 35 40
-4,0000
-6,0000
-8,0000 Time
zt Vs t
5,0000
4,0000
3,0000
2,0000
1,0000
0,0000 dd
zt
-1,0000 0 5 10 15 20 25 30 35 40
-2,0000
-3,0000
-4,0000
-5,0000 Time
32
ut Vs t
3,5
3
2,5
2
dd
ut
1,5
1
0,5
0
0 5 10 15 20
Time 25 30 35 40
Figure 4 ut = 3
10,0000
Time Serie xt
8,0000
6,0000
4,0000
2,0000
0,0000
xt
0 5 10 15 20 25 30 35 40
-2,0000
-4,0000
-6,0000
-8,0000
-10,0000
t
0,4000
0,2000
Autocorrelations
0,0000
-0,2000 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37
-0,4000
-0,6000
-0,8000
-1,0000
Lag(j)
33
Figure 7 Peridiogram of time series xt
5 Conclusion
1 Hermite’s polynomials allow us to study the probability density function
of the vNM utility inside of a n− player game. Using the approach of Quantum
Mechanics we have obtained an equivalence between quantum level and strategy.
The function of states of Quantum Mechanics opens a new focus in the theory of
games such as sub-strategies.
2 An immediate application of quantum games in economics, is in direc-
tion to principal-agent relationship. SpeciÞcally we can use m types of agents for
the case of adverse selection. In moral risk quantum games could be used for a
discrete or continuous set of efforts.
3 In this paper we have demonstrated that the Nash-Hayeck equilibrium
opens new doors to the use of entropy in game theory. Without forgetting that
the main way to evidence the Nash’s equilibria is utility maximization, we can
affirm that the human behavior arbitrates between these two elements stochastic-
utility (beneÞts) U (p(x)) and entropy (risk or order) H(p(x)). Accepting that
U (p(x))
the relationship stochastic-utility/entropy H(p(x)) is an equivalent relationship
to the ³well-known´ beneÞts/cost we present a new way to calculate equilibria:
U (p(x))
M axx H(p(x)) , p(x) : probability function and x = (x1 , x2 , ..., xn ) exogenous
variables.
4 In all time series xt , where cycles are present, it is impossible to
know the exact value of the frequency wj of each one of the cycles j. We
can know the value of frequency in a certain conÞdence interval given by wj ∈
[wj − σ w min , wj + σ w min ] where σ w min = 2σ1x . “The more precisely the aleatory
variable VALUE x is determined, the less precisely the frequency VALUE w is
known in this instant, and conversely”
34
5 Using Game Theory concepts we can obtain the equilibrium condition
in time series. The equilibrium condition in time series is represented by the
Possibility Theorem (Tables 7,8,9 and Figures 3,4,5,6,7).
ACKNOWLEDGMENT
This work was developped with Þnancial support of TODO1, www.todo1.com
and PETROECUADOR, www.petroecuador.com.ec.
The author is grateful for helpful comments to: Jorge Moyano from Diners
Club International, Gonzalo Pozo from Todo1 Services-Miami, the assistants of
8th Conference in Computing in Economics and Finance, July 2002, AIX France,
the assistants to 10th Annual Conference of the SouthEast SAS° R
User Group
USA, the assistants to M2002, the 5th annual data mining technology conference
SAS° R
2002 USA. The opinions and errors are solely my responsability. The au-
thor also thanks to Prof. Jean Louis Rulliere, GATE director, Universite Lumiere
Lyon II France, where he was Þnishing his PhD dissertation.
Corresponding Author. Fax 59322524766, email: ejimenezecu@yahoo.com
References
[1] Allais M. An Outline of My Main Contributions To Economic Science. AER,
Vol 87 No 5: 3-9, 1988.
[2] Anderson S.P., Goere J.K. Minimum effort Coordination Games: Stochastic
Potential and Logit Equilibrium. Games and Economics Behavior. 34:177-
199, 2001.
[3] Arrow K. Mathematical Models in The Social Sciences. General Systems,
29-47, 1956.
[4] Bacry H. Introduction aux concepts de la Mecanique Statistique. Ellipsis,
París,1991.
[5] Bertalanffy L. Teoría General de los Sistemas. Fondo de Cultura Económico.
USA, 1976.
[6] Binmore K. Fun and Games. D.C.Health and Company,1992
[7] Cohen-Tannoudji C., Diu B., Laloe F. Quantum Mechanics. Hermann,
París,1977.
[8] Eisert J., Scheel S., Plenio M.B. On the Impossibility of distilling Gaussian
States with Gaussian Operations. Mimeo, Imperial College of London, 2002.
[9] Eisert J. and Wilkens M. Quantum Games. Mimeo, University of Potsdam
of Germany, 2002.
35
[10] Eisert J., Wilkens M., Lewenstein M. Quantum Games and Quantum Strate-
gies, Working Paper, University of Postdam, Germany, 2001.
[14] Fudenberg D., Tirole J. Game Theory, Massachusetts, The Mit Press, 1991.
[18] Griffel, D.H. Applied Functional Analysis. John Wiley & Sons, New York,
1985.
[20] Hahn F. Stability, Handbook of Mathematical Economics. Vol II, Chap 16,
1982.
[21] Hamilton J.D. Time Series Analysis, Princeton University Press, New Jer-
sey, 1994, Chap(6).
[22] Hammerstein P., Selten R. Game Theory and Evolutionary Biology., Hand-
book of Game Theory. Vol 2, Elsevier B.V., 1994.
36
[27] Judge G., Griffiths W.E., Hill, Lütkepohl H., Lee T. The Theory and Practice
of Econometrics, John Wiley and sons., 1985.
[28] Kay, Lai Chung. (1978). A Course in Probability Theory, Harcourt, Brace &
World, INC.
[30] Karatzas I., Shreve S. Brownian Motion and Stochastic Calculus., Second
Edition, Springer-Verlag, 1991.
[31] Karlin S., Taylor H. A First Course in Stochastic Processes., Academic Press
Inc, Usa, 1980.
[41] McKelvey R., Palfrey T. Quantal Response Equilibria for Normal Form
Games, Games and economics Behavior., 10:6-38, 1995.
[43] McKelvey R., Palfrey, T.), Quantal Response Equilibria for Extensive Form
Games. Experimental Economics, 1,9-41, 1998.
37
[44] Meyer D. Quantum Strategies. Phys. Rev. Lett. 82 (1999) 1052-1055, 1999.
[45] Meyer D. Quantum Games and Quantum Algorithms, To appear in the AMS
Contemporany Mathematics volume: Quantum Computation and Quantum
Information Science, 2000.
[46] Monderer D. and Sharpley LL. Fictitious Play Property for Games with
Identical Interests., Journal of Economic Theory 68: 258-265, 1996.
[48] Neyman A. Repeated Games with Bounded Entropy., Games and Economic
Behavior, Ideal, 30, 228-247, 2000.
[55] Rubinstein A. The Electronic Mail Game: Strategic Behavior Under, Almost
Common Knowledge., American Economic Review, 79:385-391, 1989.
38
[61] Tirole J. Théorie de L’organisation Industrielle., Tome II, Economica, 1988.
[63] Weibul J.W. Evolutionary Game Theory., The MIT press, 1995.
39