Lecture Note 2007 (MS Power Point Format)

応用粒子線科学専攻
Computer Simulation Approach of
Risk Management
(リスク情報科学特論)
他 専 攻
Topics of Synergetics
(シナジェティックス特論)
Textbook and References
1. 教科書
「現代の力学 ー 線形と非線型の世界 ー」
田附・小澤 著, 学術図書出版社
2. 参考書
・ 「計算機シミュレーション入門」
小澤・へールマン 著, 学術図書出版社
・ 「若者の感性とリスク」
土田昭司・伊藤誠宏 著,北大路書房
・ 「安心安全のための社会技術」
堀井秀之 著,東京大学出版会
経験→学習(記憶・理解)
経験(実戦)
他人の経験
沢山の経験の記憶
論理的理解(定石)
感性的理解(センス)
その他: ゴルフ,書,物理学,投資(株),・・・
十分な学習後の思考: 直感精読
論理的知識: 記述(表現)できる知識(知識の伝承)
感性的習得: 記述(表現)できない知識(センス,感
性,フィーリング,・・・)
論理的思考: 「ことば」で考える(人間的・意識的)
直感的思考: 直感する(動物的・無意識・無意図的)
例: 因数分解,数式変形,スポーツ,・・・
論理的思考:対話
問い
弁証法 (Dialectic ):正(テーゼ),
反(アンチテーゼ),合(ジンテーゼ)
答え
問い
答え
問い
答え
問い
必要条件:表現(記述)の完全性
答え
・・・・・・・
・・・・・・・
科学と非科学
科学: 記述の完全性
ものごとを記述しつくすことができる。その記述に従って
ものごとが完全に再現(把握)できる。
例 ニュートン力学,(量子力学)
非科学: 記述の不完全性
ものごとを完全に記述できない。その記述をもとにものご
とが再現できない。(多くの解釈が存在)
例 スポーツ,音楽(演奏),文学(朗読),・・・
How do you describe Nature?
(いろいろな「ことば」:論理的表記法)
― Method of Sciences ―
• When you have succeeded in describing, you have
succeeded in understanding.
→ How do you describe ?
(→ How do you express what you are thinking ?)
→ How do you express how you are feeling ?
• Natural Languages
Japanese, English, French, German……
• Artificial Languages
(1) Mathematical expression → Theoretical Approach
(2) Numerical expression (Graph) → Experimental Appr.
(3) Computer code → Computational Approach
(4) Music code, Midi code → Artificial Approach (Music)
Three Categories of Scientific Approach
Computational
Simulation≒Understanding
(Computer codes)
Experimental
Theoretical
Observation
Understanding
(Numerical Data)
(Mathematical Expression)
Experimental
観察:物理量と物理
量の関係を調べる
何を視るか?
物理法則:1を知っ
てすべてを知る。
観察・思考の倹約
Galileo Galilei, 1564.2.12 – 1642.1.8
ガリレオは,なぜランプの運動
に注目し得たか?
Experimental
いかに視るか?
実験装置
いかに記述(記録)するか?
数(グラフ)⇒数学(数式)
数学は神が宇宙を書くためのアルファベットだ。
イタリア語:"La Matematica è l'alfabeto nel
quale Dio ha scritto l'universo."
Theoretical
Sir Isaac Newton, 1642.12.25-1727-3.20) 「自然哲学の数学的諸原理(1687)プリンキピア)」
「変化」を記述する「ことば」: 流率法(微積分法)
の発明
運動の法則
+
万有引力の法則
惑星(天体)の運
動の完全理解
ニュートンは,なぜ流率法(微分法)を考え
得たか?
静的情報:正確,動的情報:無
静的情報:不正確,動的情報:有
リンゴの性質(静的な性質): 色,形,大きさ,味,硬さ,位置(高さ)=x:静的な物理量
リンゴの性質(動的な性質): 色の流率(変化率),形の流率,大きさの流率,味の流率,
硬さの流率,位置(高さ)の流率(変化率):動的な物理量
静的な物理量
x
静的な物理量の流率(時間変化率)
dx
x 
dt
軌道:流率の積分(未来を知る)
m v  F ( x)
x  v
m x  F (x)
2
x(t )
微分方程式:物理量とその流率との関係
⇔ 物理法則
d x
m 2  F ( x)
dt
万有引力(重力)
ニュートンは,なぜ万有引力を考え得たか?
mM
F ( x)  G 2
x
Computational (1)
計算機で何ができるか?
計算機とは何か?
⇒ 論理的な処理
⇒ チューリング機械
チューリング機械とは何か?(論理的な処理の定義)
⇒ 無限に長いテープ,その中に離散的に格納された情報を読み
書きするヘッド,機械の内部状態(有限)を記憶するメモリで構成
される。内部状態とヘッドから読み出した情報の組み合わせに応
じて,(1)ヘッドの位置の情報を読みとる,(2)ヘッドの位置に情
報を書き込む,(3)機械の内部状態を変える,(4)ヘッドを右か
左に一つ移動する の動作のいずれかを一意的に決定でき(その
プログラムが存在する),機械の内部状態が停止状態になるまで
反復して実行する機械。
Computational (2)
現在の計算機で問題がけるとは?
⇒ そのチューリング機械を見つける(つくる)ことができること。
すべての論理的な問題は解けるか?
⇒ No! 一般にチューリング機械が停止するか否か(問題が解
けるか否か)を判定するチューリング機械が存在しないことが示さ
れている。(命題が正しいか否かを論理的に決定できない論理的
な問題が存在する。):ゲーデルの不完全性定理
計算機を考え直す ⇔ 論理的に(理屈で)結論がでない
問題(両立しない問題)の攻略法(対処法)のヒント
Computational (3)
理論的方法の補助としての計算機
 正確な実数計算ができない。
 無限に大きな数,無限に小さな数が取り扱えない。
 自然の記述(理論的): 微積分 ⇔ 実数の連続性
◆ 計算機による自然の記述:
近似的だが実用上十分な精度
計算的手法がそれ自体独自の研究手法となるためには
(思考のさらに強力な道具となるためには):
(1)アルゴリズム的自然記述のフレキシビリティ
(2)バーチャル空間での思考
(3)思考の媒体としてのプレゼンテーション法
(例)人工生命,仮想粒子,人工社会・・・・
論理的思考過程における直感の役割
問い
答え
問い
答え
問い
答え
問い
直感的思考:センスの良さ:背景的問題意識(答えの理
解の仕方,哲学,理想,価値観,人格,人生観,・・・)
答え
・・・・・・・
・・・・・・・
How to set a problem (1)
「・・・科学の方法は観測と実験とに存する。科
学者にかすに、もし無限の時間を以ってす
るならば、彼にむかってただ『視よ、しこうし
て正しく視よ。』とのみ告げれば足るであろ
う。しかしながら、科学者は一切を視つくす
ほどの時間をもたない。また、下手に視るく
らいならば、全然視ないに如かぬ。よってこ
こに選択の必要が生じる。それ故に、まず
如何にして選択をなすべきか、それを知る
のが第一の問題となるのである。これは歴
史家のみならず、物理学者にもさらにはま
た数学者にも課せられる問題であって、こ
れらの各学者を導く原理の間にはおのず
から相類似するところがないわけでもない。
科学者は本能的にそれらの原理に従って
行く。・・・」
「科学と方法」ポアンカレ著、吉田洋一訳、岩
波文庫 より
Jules-Henri Poincaré,
1854.4.29 – 1912.7.17
How to set a problem (2)
科学者が本能的に従う原理
(科学者のセンス:気持ちがいい手筋)
• 類推(アナロジー)による接続的拡張
3次元の幾何学から4次元の幾何学の構築
• 美的感覚
対称性,統一性,一元性,普遍性・・・
• 先入的信念
「神はさいころ遊びをしない」(神=創造主=自然)
• 表現形式の整備
きれいな表現の追及,記述の完結性,・・・
How to set a problem (3)
西欧的発想と東洋的(日本的)発想
ヨーロッパ文化(西欧,中欧,東欧)
= ギリシャ・ローマ文化(古典)
+ キリスト教文化(中世)
+ ゲルマン・スラブ原始文化
西欧的: 論理的(部分の積み重ねで全体を視る・段階
的)・自然対峙的 (医学,哲学:科学,芸術:絵画・音楽)
日本的: 感性的(全体を全体として視る・総体的)・自然同
化的 (漢方,わび・さび,○○道,さとり,解脱・・・)
論理的思考過程における知識の役割
問い
答え
問い
答え
問い
答え
問い
答え
・・・・・・・
・・・・・・・
How to think (1)
ヒルベルトの形式主義
ポアンカレの経験
ポアンカレの
数学的直観主義
ポアンカレはフックス関数に類似なものはないことを証
明しようとしていた。約2週間ほど、ミルクを入れない
コーヒーを飲みながら懸命に考え続けた。そしてある
夜、超幾級数から誘導されるフックス関数の一部類の
存在を証明すればいいと気づいた。彼はさっそくデー
タフックス級数を創造してみたがうまく証明が進まない。
その後多忙に紛れ、その課題を考える時間がとれな
い状態が続いた。それが,旅先で乗合馬車に乗ろうと
してステップに足をかけた瞬間に、フックス関数を定
義するために用いた変換は非ユークリッド幾何学の変
換と同じであるという考えがふと浮かんだ。これがきっ
かけとなり、問題を次々に片付け、あと一つの難関と
いうところで、また作業がストップした。このとき、兵役
が回ってきて、この数学の問題を考える時間がとれな
Jules-Henri Poincaré,
くなった。 それがある日、ある大通りを横断していると
1854.4.29 – 1912.7.17
きに、その最後の困難を突破する糸口がひらめいた。
How to think (2)
意識,自我(エゴ)
前意識
無意識(エスイド)
無意識は興奮に沸き立つ釜
Sigmund Freud
(1856-1939)
「夢判断」 1900年
How to think (3)
Linkage model in Cognitive Psychology
Stimulation (Question, etc.) → Thinking → Decision making
Linkage of stored knowledge → Logistic meaning (Decision)
Brain of Ai
Stimulation
JCO Accident
Ki(x1)
Job from JCO
Ki(x2)
Decision
Radio activity
Ki(x3)
Hiroshima
Ki(x4)
Di(x)
Note: Linkage is Stochastic in nature → Decision is also stochastic.
How to think (4)
意識,自我(エゴ)
Stimulation
前意識
無意識(エスイド)
記憶された(学習された)知識の状態がキーポイント
⇒
使う準備・整理:ラベル付け(関連付け・意味づけ)
Psychological test: Which dish do you take?
Fish from Tokai
Fish from distant
area
Decision is based on knowledge !
What impression have you got from the
video (Image information)?
An additional Message (Linguistic information) :
• An atomic energy plant is 10km from the port.
(100% true)
• The cooling water (2nd) of atomic energy plants is
drained into the sea and the temperature near the outlet
is 2.5 degrees higher :Heat Pollution.
(100% true)
• Some people say that there is a possibility that the
cooling water is contaminated by radio activity.
(quite uncertain information)
How have you changed your impression of the
see by this message?
Psychological test: Which dish do you take?
Fish from Tokai
Fish from distant
area
Decision is based on knowledge !
Long term variation (Stress annealing)
Just after
the accident
Recent
Before the accident
↑
Missing of Job
opportunity
↑
Affect to
agriculture
↑
Decreasing of
land value
Degree of threatening feeling from different ideas
The artificial society is a nonlinear system
Decision
making process
Knowledge
Decision
Message
making process
Modification of
knowledge by
message
Message
思考・発言によって,蓄積された知識が整
理される(意味付け,関連付け)
思考・発言による知識の整理(意味付け,関連付け)は社
会全体で行われ,個人にそのフィードバックがある⇒個人
だけで思考しているわけではない。
Artificial Society
Agent
Agent
Character
Knowledge
Character
Knowledge
Agent
Character
Knowledge
Channels
Agent
Agent
Character
Character
Knowledge
Message
Knowledge
思考に影響する因子
環境(孤独,人前,集団,時間条件,社会,・・・)
精神状態(不安,恐怖,緊張,・・・)
感覚情報(視覚,聴覚,温度,・・・)
蓄積知識とその状態(忘却,反すう,興味,・・・)
直感:センス:知識の結びつき方
Stimulation
JCO Accident
Ki(x1)
Job from JCO
Ki(x2)
Decision
Radio activity
Ki(x3)
Hiroshima
Ki(x4)
Di(x)
知識の管理
 記憶(収得知識)の強弱は印象の強さに比例する。印
象の強さは対象に対する注意力に比例する。注意力は
興味に関係する。(何に意を止めるか?)
 注意しないものは,生理的に感知さえできない。(生理
的フィルター)
 興味を持って最大の注意力で得た知識は,良い状態
(センス良くラベル付けされた状態,関連付いた状態)
で記憶される。
 個々の知識を単に機械的に記憶するのではなく,全体
的な枠組み(モチベーション,興味)の中で意味を持た
せて管理する。
 良い状態の知識の集合と全体的な枠組みがセンス良
い判断を生む。
まとめ
(1)1つのことに集中して懸命に考える(作業を
する)十分な時間をつくること。
(2)集中して懸命に考える(作業をする)方法を
工夫すること。
(3)答えがでない場合には、思考(その作業)
を適宜中断すること。
(4)得られた答えは絶対的でないことを知って
おくこと。
(5)常々知識の習得と整理の作業を行いセンス
を向上に努めること。(情熱・興味・問題意識
のベクトルを吟味すること)
What is Synergetics?
• Nonlinear System
→ No analytical approach
→ Computational approach
• Target System to be studied
Mechanical system, Chemical system,
Molecular system, Cosmic system, Biological
System, Social System, etc.
→ Natural Science + Social Science
• Dynamical properties as well as Static
properties of the system being studied
Three Categories of Scientific Approach
Computational
Simulation≒Understanding
(Computer codes)
Experimental
Theoretical
Observation
Understanding
(Numerical Data)
(Mathematical Expression)
Linear and Nonlinear Systems
x
y
y = f (x)
Linear
y
Nonlinear
y
y = ax + b
y = ax
x
x
• Superposition Principle holds
• Superposition Principle does not hold
F(x1) = y1, F(x2) = y2 → F(x1+x2)=y1+y2
F(x1) = y1, F(x2) = y2 → F(x1+x2)≠y1+y2
Result is Predictable !
Result is Non-Predictable !
Method
1. Computer Simulations and Analysis :
Artificial Society Model (Monte Carlo
Simulation) , Numerical Linguistics
2. Experimental (Experimental Psychology):
Questionnaire Experiments, Classroom
Experiments, Online Network Experiments
3. Theoretical : Social Psychology
(Communication Psychology), Cognitive
Psychology, Theory of Numerical Linguistics,
Syntax Theory, Theory of Panic, Risk
Management Theory
Related Topics
1. Order and Disorder Problems in Statistical
Mechanics (Monte Carlo Simulation of propagation
of order and disorder : Kinetic Ising System
(Temperature Barkhausen Effect, AFT
Lattice=Frustrated System) , Molecular System
(Amorphous Ice)
2. Acceptability of Commercial Product and
Information Given to Customers (Questionnaire
Experiments) : Communication Psychology,
Statistical Linguistics, Computer Numerical
Linguistics
3. Artificial Society Model (Monte Carlo Simulation
of Decision Making in Society) : Simulation of
Panic, Risk Management
Let us start with an experimental approach
Rigid Body Pendulum
• How do you observe the
motion of a double
pendulum ?
• How do you record it?
• How do you measure it?
How about an analytical approach?
(Computational approach is based on it)
• How do you describe the motion of a double
pendulum ?
• Modeling
Point Mass Pendulum
Rigid Body Pendulum
L
L1
m
L
I1
L2
I2
m
M2
M1
Do you have knowledge of Analytical
Mechanics ?
Yes: Try to solve the problem
No: Read the textbook at pages 12-20,
and 192-198
Exe.1 Single Pendulum
Kinetic Energy: (½ ) m (L dθ/dt)2
Potential Energy: - m g L cosθ
Lagrangian :
L
L = (½ ) m (L dθ/dt)2 + m g L cosθ
θ
Equation of motion:
m
→ see at p.19 of the textbook
How do you solve the equation of
motion ?
→ see at p.19 of the textbook
Exe.2 Double Pendulum
Kinetic Energy: ?
Potential Energy: ?
L
θ1
Lagrangian :
m
L=?
L
Equation of motion:
θ2
m
→?
How do you solve the equation of
motion ?
→?
Exe.3 Single Pendulum
Kinetic Energy: (½ ) m (L dθ/dt)2
Potential Energy: - m g L cosθ
Generalized Momentum :
L
p=
θ
∂ L /∂(dθ/dt)
Hamitonian :
m
H = 1/(2 m) p2 - m g L cosθ
Equation of motion: ?
How do you solve the equation of
motion ?
Report: Problem 1
L
θ1
m
L
θ2
m
• Show the Hamilton’s
function for the double
pendulum system.
• Show the equation of
motion.
• Explain how and why the
system becomes chaotic.
• Write the report in A4
format and present it before
23/07/2007 (Hitachi) or
06/08/2007 (Mito)
Dynamics of System (see Textbook at p.37)
Hamiltonian : H (q1,…qf, p1,…pf)
Equations of Motion :
dqi/dt = ∂ H / ∂pi
dpi/dt =
-∂
→
H / ∂qi →
dqi = (∂ H / ∂pi ) dt
dpi =
-(∂
H / ∂qi ) dt
New State (q1,…qf, p1,…pf)new
New State
Old State
= Old State (q1,…qf, p1,…pf)old
+ Displacement (dq1,…dqf, dp1,…dpf)
2f-Dimensional Phase Space
Let us begin with an demonstration
• Simple mechanical structure
→ Chaotic motion
How to observe the motion precisely?
(Modern way of observation)
• Monitor the movement by digital video camera
• Analysis the image analysis on PC
.
hinge with red marker
upper bar
hinge with green marker
lower bar
red marker
Give enough experimental information (1)
(Measurement of friction force at hinges)
Time (sec)
friction force = constant
Give enough experimental information (2)
Fair (N)
(Measurement of air resistance force, Fair)
Air stream velocity, v
(m/s)
Fair = kv2
Express analytically as precisely as possible
(Easy simplification or modeling is no good!!)
• Lagrangian formulation
2 2
1 2 1  2 1
2 2
L  I11  I 2 2  m2{l1 1  h2 2  2 l1h212 cos(1   2 )}
2
2
2
 m1 gh1 cos1  m2 g (l1 cos1  h2 cos 2 )
• Dissipation function
K  K1  K2
K2 
K1  C1 | 1 | C 2 | 2 |
1
l
1
D1{1  2 | cos(1   2 ) |}3 | 1 |3  D2 | 2 |3
3
l1
3
• Equation of motion
d  L  L K

 
  0

dt  i  i i
(i  1, 2)
Everything through computer simulation
• Numerical integration of equation of motion
( Runge-Kutta method)
→ Produce all the possible motions in computer
• Graphics presentation
(Original graphics software, SGL)
→ Interactive study with computer
What to see is determined by students!!
To see “pure chaotic motion” in computer
(Chaos is easily veiled by energy dissipation
→ neglect energy dissipation terms)
Quasi-harmonic
Chaotic
Lower bar
Energy (J)
Upper bar
Sum of the two
Time (sec)
(a) low energy initial state
1   2  40, 1  2  0
Time (sec)
(b) high energy initial state
1   2  110, 1  2  0
Poincare’s mapping
Quasi-harmonic
Angular velocity (rad/sec)
Chaotic
1
(rad)
(a) low energy initial state
1   2  40, 1  2  0
1
(rad)
(b) high energy initial state
1   2  110, 1  2  0
Fourier analysis
Quasi-harmonic
Chaotic
(b)
(a)
Frequency (Hz)
Frequency
(Hz)
(a) low energy initial state
1   2  40, 1  2  0
Frequency
(Hz)
(b) high energy initial state
1   2  110, 1  2  0
Quasi-harmonic
“Pure” chaotic
Upper bar
Energy
(J)
Lower bar
Sum of the two
Time (sec)
Time (sec)
Energy (J)
Harmonic with energy dissipation Chaotic with energy dissipation
Time (sec)
Time (sec)
“Pure” chaotic
Angular velocity (rad/sec)
Angular velocity (rad/sec)
Quasi-harmonic
1
1
(rad)
Chaotic with energy dissipation
Angular velocity (rad/sec)
Harmonic with energy dissipation
(rad)
1
(rad)
1
(rad)
“Pure” chaotic
Quasi-harmonic
(b
(a
))
Frequency
(Hz)
Frequency (Hz)
Frequency (Hz)
Chaotic with energy dissipation
Power spectrum
Harmonic with energy dissipation
Frequency (Hz)
Frequency (Hz)
Why Chaotic Motion occurs in the
double pendulum system?
In order to answer this question, let us see
our textbook at pages 181-186.
The double pendulum system and Heron &
Heiles system are classified into the chaos
system with small number of freedom of
motion
Dynamics of System (see Textbook at p.37)
Hamiltonian : H (q1,…qf, p1,…pf)
Equations of Motion :
dqi/dt = ∂ H / ∂pi
dpi/dt =
-∂
→
H / ∂qi →
dqi = (∂ H / ∂pi ) dt
dpi =
-(∂
H / ∂qi ) dt
New State (q1,…qf, p1,…pf)new
New State
Old State
= Old State (q1,…qf, p1,…pf)old
+ Displacement (dq1,…dqf, dp1,…dpf)
2f-Dimensional Phase Space
Report: Problem 2
• Show the Hamilton’s function for the HeronHeiles system.
• Show the equation of motion.
• Explain how and why the system becomes
chaotic.
• Write the report in A4 format and present it
before 23/07/2007 (Hitachi) or 06/08/2007
(Mito)
Deterministic to Stochastic
Nonlinear System
• Chaotic
• Ergodic
⇔
Trajectory Mixing
• Liouville’s Theorem
Assumption: (等しいア・プリオリ確率の仮定)
Equal probability of appearance of states of
an equal energy
Trajectory of Ergodic System in Phase Space
Note :
No crossing in
a trajectory!
Plane of Equal Energy
2f-Dimensional Phase Space
Trajectory of Chaotic System in Phase Space
Note :
No crossing of
two trajectories!
Plane of Equal Energy
2f-Dimensional Phase Space
• Liouville’s Theorem → Any volume in the phase space is conserved during the
time development of the system.
• Chaotic system → Distance between any two points in the phase space increases
exponentially with time.
How the trajectory is mixed
Can you have an image of the trajectory of
both chaotic and ergodic system?
The trajectory would be highly complicated.
The trajectories would be highly mixed.
No need to trace the deterministic time
development of the system and just obtaining
stochastic information is enough.
Assumption: (等しいア・プリオリ確率の仮定)
Equal probability of appearance of states of
an equal energy
Probability of appearance of different
energy states
Let us consider a system of N particles of a same kind but
they are treated as different particles (classical particles).
Each particle can have one of possible states.
P1
ε1
Pk
…
…
ni
…
ε2
P3
…
… …
εi
P2
n2
n1
A state of the system (ε1ε3ε8 . . . εi . . . )
Note: These are particle states
presented by its energy state
….
Number of cases W in which n1 particles in state ε1,
n2 particles in state ε2, … ni particles in state εi …
N!
W
n1! n2!  ni ! 
Probability of occurrence of a distribution (n1, n2, … ni,
…) is proportional to the number of cases W (n1, n2, …
ni, …) because the each states have an equal probability
of appearance.
When N is large, the function W (n1, n2, … ni, …) has
very sharp peak, and almost all states belong to the
distribution of the peak.
In actual cases, only the distribution of the peak appears.
Let us look for the maximum of W
N!
W
n1! n2!  ni ! 
Where
n
i
i
N
 n
i
i
E
i
ni  A exp(    i )  A e
  i
Boltzmann factor : exp(    i )  exp(  i / kT )
Lagrange Multiplier (ラグランジェの未定定数法)
Example Determine the extremal values
(maximum or minimum values) of f(x,y)=xy
under the condition x2+y2-1=0
F ( x, y )  f ( x, y )    (boundary condition )
 xy   ( x  y  1)
2
2
F
 0  2 x  y  0
x
x?
F
0 
y
xy  ?
B. C
x  2y  0
x2  y2 1  0
y?
 ?
Report: Problem 3
Determine the distribution (n1,n2,…ni,...) which gives
N!
Maximum of W 
n1! n2!  ni ! 
under the following
boundary conditions.
n
i
N
 n
i
i
i
i
Hint
Use Stirling’s formula : log n!=n log n - n
And look for the maximum of logW
E
Idea of Gibbs
Pressure
Mean Value of Physical
Quantity of a System:
Time
Time Average
Average over ensemble of
microscopic states which
belong to a macroscopic state
Josiah Willard Gibbs
(1839 – 1903)
Statistical Mechanics
Micro Canonical Ensemble: Isolated System
Number of particles N=const.
Total energy of the system E=const.
Canonical Ensemble: System of a Definite Temperature
Number of particles N=const.
Grand Canonical Ensemble:
System of a Definite Temperature and Chemical
Potential
Trajectory of Ergodic System (Micro Canonical)
Isolated
Plane of Equal Energy
2f-Dimensional Phase Space
Trajectory of Ergodic System of A Defined
Temperature (Canonical )
Heat Bath
T=const.
Energy Fluctuation
E0  E  E0  E
2f-Dimensional Phase Space
Trajectory of Ergodic System of A Defined
Temperature and Defined Chemical Potential
(Grand Canonical )
Particle
Source
Heat Bath   const.
T  const.
Energy Fluctuation + Dimension Change
E0  E  E0  E
2f-Dimensional Phase Space
Production of Microscopic States (1)
Deterministic Method : Molecular Dynamics (MD)
Equation of motion & a set of initial values
Hamiltonian : H (q1,…qf,
p1,…pf)
Equations of Motion :
dqi/dt = ∂ H / ∂pi
= (∂ H / ∂pi ) dt
→ dqi
dpi/dt = -∂ H / ∂qi →
= -(∂ H / ∂qi ) dt
dpi
2f-Dimensional Phase Space
Production of Microscopic States (2)
Stochastic Method : Mote Carlo Method (MC)
Markov Chain & a set of initial values
Markov Chain
Stochastic development of states
x0  x1  x2      xn    
P( x|
n x n 1 ,    x1 , x 0 )  P ( x|
n x n 1 )
MC Methods for Different Ensembles
Micro Canonical Ensemble MC: シミュレーション物理
学,シュプリンガー東京,小澤・篠島 訳 P.81
Number of particles N=const.
Total energy of the system E=const.
Canonical Ensemble MC:
同 P.86
Temperature T=const. Number of particles N=const.
Grand Canonical Ensemble MC: 同 P.104
System of a Definite Temperature and Chemical
Potential
Probability Pr( i ) that the particle Pk takes
an energy state  i
P1
P3
P2
ε2
… …
… …
…
…
i
εi
Pk
ni
n2
n1
ε1
Micro Canonical
ni  A e
  i / kT
e
  i / kT
Canonical
Pr( i )  e
  i / kT
Canonical Ensemble MC Method
(1) Define an initial state, x 0
(2) Generate a state x by using R. N.
(3) Calculate the energy change, H associated
with the transition x 0  x
(4) If H is negative, accept x and go to (2)
(5) If H is positive, calculate exp(  H kT )
(6) Generate R.N. R  [0, 1]
(7) If R  exp(  H kT ) accept x and go to (2)
(8) Otherwise, accept x 0 and go to (2)
Ising Spin System (Static and Kinetic)
H  -J  si s j
i, j
si , s j   1
J  0 : Antiferomagnetic
J  0 : Feromagnetic
Let us see demonstrations in the reference
FPU Problem
Look at our textbook at pages 142-166.
• Fermi-Pasta-Ulam, 1955
• Mechanical explanation of Heat and Sound
• Ergodic system
• Computer Simulation Method
• 1D Mass-Spring Lattice with P.B.C
1
2
H ( q1 ,  , q f , p1 ,   p f ) 
pi

2m i
c (1)

2
( 2)
( 3)
c
c
2
3
(
q

q
)

(
q

q
)

i i1 i

i 1
i
3 i
4
4
(
q

q
)
 i1 i
i
Nuclear accident in
1999 at JCO Tokai
• Chain nuclear reaction
of Uranium started by
mistakes during
chemical processing of
atomic plant fuel
• Two JCO peoples died
of being exposed to the
radiation
• No serious damage to
31 thousands of people
living within 10km
circle region
Neutron and γ- ray data at two monitoring points
Neutron data peak :
The time of accident
Psychologist group of
Ibaraki University
investigated how the
public opinion
concerning atomic
energy plants changed
due to the accident
How people got the first
news of the accident?
How people got the detailed
information of the accident?
People’s talk 10-20%
People’s talk 5-10%
TV 60-70%
TV 70-80%
Main Source of information : TV + Newspaper
Short term change in
public opinion
Before
After
Atomic energy plants
are safe?
Positive answer :
60% → 15%
Do you support atomic
energy plants?
Positive answer :
80% → 50%
Before
After
Long term variation (Stress annealing)
Just after
the accident
Recent
Before the accident
↑
Missing of Job
opportunity
↑
Affect to
agriculture
↑
Decreasing of
land value
Degree of threatening feeling from different ideas
What impression have you got from the
video (Image information)?
An additional Message (Linguistic information) :
• An atomic energy plant is 10km from the port.
(100% true)
• The cooling water (2nd) of atomic energy plants is
drained into the sea and the temperature near the outlet
is 2.5 degrees higher :Heat Pollution.
(100% true)
• Some people say that there is a possibility that the
cooling water is contaminated by radio activity.
(quite uncertain information)
How have you changed your impression of the
see by this message?
Psychological test: Which dish do you take?
Fish from Tokai
Fish from distant
area
Decision is based on knowledge !
Artificial Society
Agent
Agent
Character
Knowledge
Character
Knowledge
Agent
Character
Knowledge
Channels
Agent
Agent
Character
Character
Knowledge
Message
Knowledge
Artificial Society Model
• ARTIFICIAL SOCIETY is made of various AGENTS.
• The AGENTS represent various organizations of a real society, i.e. individual
persons, mass communications (broadcast stations, press companies, internet
media sources, etc.), commercial companies, government, local governments, etc.
• The each AGENT has a certain KNOWLEDGE related to a given PROBLEM.
• The each AGENT makes a DECISION of the given PROBLEM based on the
KNOWLEDGE.
• The AGENTS have CHARACTERS different each other in decision making.
• There are CHANNELS between AGENTS.
• AGENTS make MESSAGES on the bases of their DECISIONS and their
KNOKEDGE.
• The each AGENT has an individual CHARACTER in message making.
• MESSAGES are transferred through definite CHANNELS between AGENTS.
• AGENTS change their KNOWLEDGE as the results of MESSAGE exchanges.
The artificial society is a nonlinear system
Decision
making process
Knowledge
Decision
Message
making process
Modification of
knowledge by
message
Message
How much psychological lows are taken into
account ⇔ How much the model is realistic.
Elements of the model
Agents : Ai i=1,2, …. N
i=1 Citizens (People with average knowledge)
i=2 Aggressive group (Green Peace, …)
i=3 Mass media (TV, Newspaper, Publisher , …)
i=4 Government (Central and/or Local, Foreign countries )
i=5 Political party (Party A, Party B, …)
i=6 Atomic energy plants company (Employer, Employee)
i=7 Scientists (Social scientist, Natural scientist)
• Agents do not change their character
• Neglect appearances of new agents
→ Artificial society model is static in structure
Knowledge of Ai concerning a given problem: Ki (x)
X : Parameter (label) to a certain knowledge
Ki(x) : Intensity of the knowledge, x
Cf
ψi(x) : Quantum mechanical state of molecule, i
Example: Atomic energy plants are safety?
X
: Degree of safety
(x=1 : 100% safety)
Knowledge structure related to
negative answer
Knowledge structure related to
positive answer
Ki(x)
Wide range knowledge
0
1
x
Let us consider a linguistic message
Message from Ai to Aj : Mij(x)
A message is made of many information
(sentences) .
X : Parameter (label) to certain information
Here, we use the same x as in Ki(x)
Mij(x) : Intensity of the information, x
Mij(x)
0
Ki(x)
1
x
Decision making process
Stimulation (Question, etc.) → Thinking → Decision making
Linkage model in Cognitive Psychology
Linkage of stored knowledge → Logistic meaning (Decision)
Brain of Ai
Stimulation
JCO Accident
Ki(x1)
Job from JCO
Ki(x2)
Decision
Radio activity
Ki(x3)
Hiroshima
Ki(x4)
Di(x)
Note: Linkage is Stochastic in nature → Decision is also stochastic.
How do you express it mathematically?
Select xi randomly
with weight Ki(x)
Ki(x)
x1 x2 xN
x1: Hiroshima
x2: Radio activity
xc: Conclusion
xN: Job from JCO
Di(x)=ΗKi(x), Η: Operator
Xc=ΣxjKi(xj)/
x
ΣKi(xj)
σ2=Σ(xj-xc)2/N
Ki(xc)= ΣKi(xj) /N
Xc
x
Gaussian
Function,
Di(x)
Message formation process
Stimulation (Question)
→ Thinking (Decision making) Di(x)
→ Message Mij(x) = Cij(x)・{Di(x) + Ki(x)}
Examples of characteristic function Cij(x)
Simple minded (Express as it is) : Cij(x) =1
Authority dependent : Cij(x) =1 for J=Authority
Cij(x) =0 for J= Non-authority
Di(x)
Ki(x)
x
Knowledge modification process
• When receiving a message Mji(x) from Aj
K’i(x) = Cij(x)・Mji(x)
Cij(x) : Characteristic function of Ai (filtering function of Mji(x))
Examples (Authority effect in Social Psychology)
Simple minded (Accept message as it is) : Cij(x) =1
Authority dependent : Cij(x) =1 for J=Authority
Cij(x) =0 for J= Non-authority
Self dependent : Cij(x) = Ki(x)
• Variation with time (Sleeping effect in Cognitive Psychology)
K”i(x) = Ti(x)・K’i(x) + Ki(x)
Examples (Sleeping effect): Ti(x) > 1 for unreliable information
Ti(x) < 1 for reliable information
Summary
• Rough idea of public opinion formation model
• The work is just started (code checking level) and no
accumulation of data
• Most important is the role of mass-communication and
that of government.
• How these two agents contribute to the formation of
public opinion is my interest.
• If some rules for the problem are found from my
computer simulation study, the work would be
successful.
• Welcome to your kind collaboration to my study.
Report: Problem 4 (Questionnaire)
(1) パワーポントの資料(教員のHP)をオンラインで見たか? YESの場合,どこで見たか?
(2) パワーポイントの資料をダウンロードしてPC上で見たか? YESの場合,どこで見たか?
(3) パワーポイントの資料を印刷して見たか?
(4) パワーポイントの資料は理解を深める上で役に立ったか?
(5) パワーポイントの資料を何回見たか?
(6) 授業のビデオ(教員のHP)をオンラインで見たか? YESの場合,どこで見たか?
(7) 授業のビデオをダウンロードして,PC上で見たか? YESの場合,どこで見たか?
(8) 授業のビデオは理解を深める上で役に立ったか?
(9) 授業のビデオを何回見たか?
(10)教科書を購入したか?
(11)教科書は何回見たか?
(12)教員のHPの参考文献をオンラインで見たか?
(13)それ以外に自分で調べたか? YESの場合,何をどのようにして調べたか?