[제 2회 jsc afterschool] 모델과 시나리오 (유병수, 130810)
DESCRIPTION
모델이란 무엇인가, 그리고 모델이 왜 중요한가 모델은 어떻게 만들어지는가에 대해 다룬 자료입니다TRANSCRIPT
Model and Scenario
Byeongsu Yu
Yonsei, Univ.
Contents
• Introduction
• Definition
• How to make model
• Exemplary– Dim 1 : Abstract/Concrete
– Dim 2 : Common Sense/Systemized
– Dim 3 : Similarity of methodology
– Dim 4 : How disciplinarys can be made?
• Conclusion
Story Line
• Introduction
• Showing actors and basis storyline
• How to make story
• Climax or Climaxes?
• Variation 1
• Variation 2
• Variation 3
• Variation 4
Introduction
• Fired Theater model
Introduction
• Fired Theater model
Introduction
• How about this?
– 기둥을 중간에 설치하면 어떻게 될까?
– 사람들이 가는 속도가 오히려 더 빨라진다!
Definition
• Model
– Depict phenomenon
– Space and Relationship
• Scenario
– What can occur in “Real World”
– Realized path
Definition-Model
• Space
– Sets of assumption
• Actors, Outcomes, Actions, Things, etc.
– Is it measurable? (or well-defined?)
• Finite, Countable, Uncountable
– How many dimensions it have?
• Dimension is number of sets in space
– Can we make disjoint set?
Definition-Model
• Relationship
– Laws, Theorem, etc.
– How spaces can related with each others
• Set of n-tuples
– Is it function?
– Transitive, Symmetric, reflexive.
Introduction
• Why scenario matter?
– Because it shows what we can’t imagine easily, especially making model.
– Ex: Prey-Predator
• 풀, 양, 늑대
– Ex : So many historical examples
• The Great depression, Maginot line, etc.
Introduction
• Why model matter?
– Though, still models are best approximation of future.
– The diversity prediction theorem
• Collective Error = Average Individual Error-Prediction Diversity
Introduction
• Why model matter?
How to make model?
• There is a coin. We know P(Head) = P(Tail) = ½. We experimented 100 times, and All the observations are head.
• You have to put your money on this game. Which side you want to choose?
– Head
– Tail
– Not both (Standing on)
How to make model
• Bayesian Approach
– X is observation. W_i = signal(Law)
– Posteriori = likelihood*Prior/evidence
How to make model
• Meaning of bayesian approach
– Depict humans approach of predict something
• Prior can help us to predict distribution
– We don’t even know parameters
• Parameter is critical value of relationship
• Actually, Bayesian approach assumes that parameter follow normal distribution, which is consistent with the law of large number
– Making model with few data
How to make story
• Making a novel
– It shows real worlds concern, which we cannot imagine easily
• Making Simulation
– As we see above.
Exemplary
• Abstract v. Concrete
– Model always delete real world’s thing
• It want to know two thing’s relationship
– However, it makes really concrete process if we combine models
Exemplary
• Common sense
– Analogy : root of Model and scenario
• Both are mixed in analogy
• Very easy to understand
• Actually, model is just another analogy to strictly logical things.
– 개똥철학
Exemplary
• 개똥 철학
– 장 뤽 고다르 – 내멋대로 해라
– 장근석… ㅠ_ㅠ
• Not all people know analogy.
Exemplary
• Humanities
– Literature
• Novels, awesome novels.
• Laws
– 오구라신페이(小倉進平) <향가, 이두에관한논문>
– 양주동 <‘청구학총’에실린원왕생가에대하여> 균여전의11수에나온한역을가지고삼대목의실전으로인한삼국유사에남은 14수를해석
– 조윤제 <시가의구체결정법칙 >
» 반절성
» 전절대후절소
Exemplary
• Humanities
– History
• E.H.Carr vs. John Lewis Gaddis
• History always explain particular cause and effect, but they do not claim that it is general law of the world
• Historian always recognize that their matters are real world’s all viewpoint things, which have infinitely many dimension and uncountable sets.
Exemplary
• Philosophy
– Philosophy : deal with linguistic concept and its property with linguistic property
• It is difficult to make disjoint concept
• Dimension of things are uncountable and infinite dimension
• Also difficult to measure
Exemplary
• Similarity of Methodology
– Science : move from philosophy to mathematics
• Economics : Aristotle, Keynes and Hayek -> Arrow– Needless to say
• Politics : Platon and Marx -> Dahl – Incentive model of party system
• Sociology : Durkheim, Marx, Weber -> Yonghak Kim– Network yeah!
• Psychology : We are not social science, idiot!– Brain works!
Exemplary
• Similarity of mathematics in social science– Actor’s or Societies optimization problem
• Difference of assumptions in social science– Economics : Incentive, All model.
– Politics : Incentive, Party model.
– Sociology : (Incommensurable) Incentive, Network model.
– Psychology : We people don’t want to bother with your social something, jerk!
Exemplary
• Similarity of philosophy with Natural Science
– Superstring things in physics
– Inference how things to be done like this world
• Make physical/Biological inference from WWII.
• Absent of measurable model in Evolutionary Science
• Difference of other science club
– Almost all things can be measurable.
Exemplary
• Practical displinary– Law
• Okay, you can see me philosophical things, but we are not philosophy because we always deal with disjoint concept which is determined by the Supreme Court!
• Also we capture every things from other disciplinary we want to justify.– Coase theorem in economics
– Deconstructionism in Philosophy
– Temporary Insanity from Medical school
– Etc.
Exemplary
• Practical Law
– Business
• Actually, we are not practical law. We are pure law, because many pure scientists come to me to do something. Hahaha, money talks!– Information Retrieval and Pattern recognition in Accounting
– Actuarial Science
– Finance : Please call me Stochastic Calculus, not business.
– Strategy : Porter –> Christenson
– Marketing : Call me Quant marketing.
– Mgmt Science : We even are not business, idiot!
Exemplary
• Et cetera
– Culinary
Exemplary
• Summary
– All models have two parts;
• Measurable part and not or difficult to measurable part
– Method is similar between sciences
– So what really matter is find the line between measurable set and unmeasurable set
• Something we dared not to measure can be measured by many tools. – CERN, Microscope, Flow, etc.
Making Disciplinary
• How we can make these method?
– Model of making model
• There are three scenario for model of making model
– Falsifiability (Popper)
– Revolution! (Kuhn)
– Proof and Refutation(Lakatos)
Making Disciplinary
• Falsifiability
– First of all, all theories have to be falisifiable. Namely, It can be false.
– Secondly, for getting approach to truth, we have to alter prior theories with posterior theory which have more explanatory power.
– Actually, Popper suppose Falisfiability to solving problem of demarcation
Making Disciplinary
• Falsifiability
– In my point, falsifiability means we have to deal with models on only measurable set.
– Also, what we have to do is comparing between models in point of sets and relation sets.
Making Disciplinary
• Problem of Falsifiability
– Without loss of generality, our first model is randomly chosen and adopting new model is continuously differ.
– Then, we can find a relation between model and its explanatory power contiuously on infinite set.
– If this relation is not linear, our maximum quantity of truth is lower than absolute maximum quantity.
Making Disciplinary
• No free lunch Theorem– From viewpoint of local maximization, finding
model is finding optimal solution for truth.
– Wolpert and Macready proved that “any two optimization algorithms are equivalent when their performance is averaged across all possible problem.”
– It says we have to adopt one model with continuously mapping, its explanatory power must be reduced on some problems.
Making Disciplinary
• Scientific Revolution
– Paradigm : universally recognized scientific achievements that, for a time, provide model problems and solutions for a community researchers
• What is to be observed and scrutinized
• Questions that are supposed to be asked
• How these questions are to be structured
• How interpreted, experimented
Making Disciplinary
• Scientific Revolution
– Normal Science
– Scientific Revolution
– (Altered) Normal Science
• Changing Model is not gradual process; It is radically changed.
• Also, two models are incommensurable;
Making Disciplinary
• Problem of Kuhn
– Relativism; Then how we can choose one model with another model? Just do what others do?
– Are these models really incommensurable?
Making Disciplinary
• Proof and Refutation
– First suppose some conjecture exists
– Three types of refutation
• Global refutation without local refutation
• Local refutation without Global refutation– Making Monsters!
– Excluding Monsters.
• Global and also local refutation– Change conjecture with more general version.
Making Disciplinary
• Sketch Book for v-e+f =2
– Imagine God as a math teacher and Geniuses (Euler, Gauss, Cauch, etc.) as his/her students on math class.
Conclusion
• Don’t exclude persons who have different model. But exclude persons without agreeing that comparing models with you on decision making.
• Do simulation or imagine a lot of scenario. Even if we have models, we actually don’t know how the model works. Even if these things are messy, please read novels.
Conclusion
• Independent n model’s is better than one fancy model. There is no free lunch on model.
– So modulation really matter on your organization or your personal decision.
– Make your own team for getting better decision
• Learning is different with knowing. Please have a proficiency on your selected model
– Proficiency really matter on blink.
Conclusion
• Model and scenario are equally important. Model says on structures we cannot see with limited information, but scenario says on plausible things with full information but without structure.
– Always we have to concern both.
The End and QnA