当前位置:网站首页>Don't know a little statistics, "Star Wars" for nothing
Don't know a little statistics, "Star Wars" for nothing
2022-07-22 02:59:00 【turingbooks】
A long time ago , In a distant galaxy ……
Our hero · Solo is driving “ Millennium Falcon ” No. 1 dodged the pursuit of Star Destroyers and titanium fighters at full speed .
In a flurry , robot C-3PO The tip says ,“ Millennium Falcon ” The superluminal thruster of No. 1 is broken . What's worse is , The spacecraft was also hit by an asteroid .
This hit Han · Soro inspiration . I saw him sitting in the driver's seat , Adjustment status , take “ Millennium Falcon ” The No. 1 went straight into the asteroid belt .
Always knowledgeable C-3PO Hurriedly prompted behind him :“ chief , The probability of successfully crossing the asteroid belt is about 1 Than 3720!”
“ Never mention probability to me !” han · Solo replied .
《 Star Wars: : Empire counterattack 》 Still
( source :imdb.com)
If seen 《 Star Wars: : Empire counterattack 》, Then you must have fresh memories of the above scenes . Behind the gripping plot , In fact, there is an interesting statistical knowledge hidden . Let's take a closer look at .
One side , We all know Han who has the aura of the protagonist · Solodine can successfully cross the asteroid belt . On the other hand , We also know that C-3PO The data is reasonable , Because no titanium fighter has successfully crossed the asteroid belt .
han · Solo thinks he has the probability of successfully crossing the asteroid belt = 100%
C-3PO The calculated probability of successfully crossing the asteroid belt = 1 / 3720 = 0.027%
Since both sides make sense , Then why is the difference between the two probabilities so large ?
likelihood :C-3PO The way of reasoning
C-3PO There's nothing wrong , It just forgot the important information —— han · Solo is a master , Super powerful !
C-3PO There is no nonsense , It is proficient 600 More than 10000 communication modes , This requires a lot of data support . So we can basically determine ,“1 Than 3720” The statement of is supported by data .
We according to the C-3PO Data given , Use β Distribution to determine the probability of success . Why β Distribution , Because given the success and failure ratio , It can correctly simulate various possible probabilities of an event .
β The parameters of the distribution are α( The number of successes observed ) and β( Number of failures observed ):
Beta(α,β)
This distribution tells us , According to the current data ,C-3PO What is the most likely probability of successfully crossing the asteroid belt .
To find out C-3PO The way of reasoning , Here are some reasonable assumptions . We assume that C-3PO Only two people have successfully crossed the asteroid belt , While the rest of the 7440 People ended their journey to the asteroid belt in a violent explosion .
The following figure shows β Distribution , namely Beta(2, 7440).
The hope of success is so slim ! No wonder Lian Han · Solo himself thinks it's dangerous to rush into the asteroid belt , So he said :“ They must be crazy to follow us .”
transcendental : han · The power of solo
C-3PO The problem with the analysis of is , It has data about all pilots , Dan Han · Solo is not an ordinary pilot .
Our transcendental belief is , han · Solo is sure to successfully cross the asteroid belt , Because he has the aura of the protagonist .
The probability of a priori belief is often controversial . A lot of people think “ Fabricate ” A priori probability is not objective . But you need to know , It is more absurd to deny a priori probability .
We have many reasons to believe that Han · Solo will survive , But there is no data to support this belief . First, let's see the upper bound of his severity .( Yes , There should be an upper bound , isn't it? ? If we believe that he will never die , Then this movie is old-fashioned and boring .)
We have a strong belief that Han · Solo will successfully cross the asteroid belt . Here again, make reasonable assumptions . Suppose our belief that he will survive is 20000 Than 1.
The figure below shows that we believe in Han · A priori probability distribution that solo will survive , namely Beta(20000, 1).
Posttest : How writers create suspense
We have made sure that C-3PO Belief ( likelihood ), Also established the Han · Thoreau's belief model ( A priori probability distribution ). Next, we need a way to combine the two , Form the final posterior belief .
Let's combine the above two as follows β Distribution , A posteriori probability distribution is obtained .
Insert specific values , Yes :
Beta(2+20000,7440+1) = Beta(20002,7441)
The posterior probability distribution can be seen intuitively from the figure below .
You can see , By way of C-3PO Belief and “ han · Solo is great ” Combine your beliefs , We found that the probability of his successful survival is about 73%.
This means that we still think Han · Solo has a good chance of success , But we will still worry about him .
Conclusion
Through this example , We learned that likelihood 、 The relationship between a priori and a posteriori . actually , These three are Three elements of Bayesian theorem .
C-3PO The data provided with us on Han · Thoreau's cognitive ability does not match the likelihood , But we did not directly deny C-3PO The data of , Instead, we combine likelihood with our prior beliefs , Form the final belief after adjustment ( Posttest ).
In the movie , This uncertainty is crucial to creating tension .
in real life , Master the basic logic of Bayes theorem , You can also use it to analyze any uncertain situation you encounter .
new book PUSH commend
边栏推荐
- Introduction to pytorch - Implementation of simple linear functions
- bootloader系列二——arm920t--bootloader架构设计
- 卷积核扩大到51x51,新型CNN架构SLaK反击Transformer
- Codeforces round 578 (Div. 2) C - round corridor [number theory + Law]
- 中断服务程序(ISR)应遵循的高效原则
- 为配置SSL,自制CA证书
- 2019杭电多校 第九场 6684-Rikka with Game【博弈题】
- unity3d-EventSystem(事件)
- quartz简单用法及其es-job
- 2019杭电多校 第六场 6641(原1008) TDL(规律题)
猜你喜欢
Introduction to pytorch II use pyplot to dynamically show the function fitting process
2022全球开发者薪资曝光:中国排第19名,平均年薪23,790美元
duilib实战1-模仿百度网盘登录界面
from scratch implement crnn using pytorch :读取训练数据
tslib-1.4移植到mini2440开发板
2019杭电多校 第五场 6630(原1007) permutation 2 (斐波那契数列)
Codeforces Round #578 (Div. 2) C - Round Corridor 【数论+规律】
简单总结一下图像处理中概念
It takes only 10 pictures to figure out how the coupon architecture evolved!
弹性体模拟(弹性力学)
随机推荐
WDK开发入门1-基础环境搭建和第一个驱动程序(VS2010)
pytorch入门二 使用pyplot动态展示函数拟合过程
bootloader系列二——arm920t--bootloader架构设计
行业分析| 物流对讲
Is redis single threaded or multi-threaded
Codeforces Round #579 (Div. 3) C - Common Divisors 【数论】
Bootloader series I -- Analysis of ARM processor startup process
树莓派3B搭建Flink集群
870. 约数个数
2019牛客暑期多校训练营(第六场)D-Move 【暴力枚举】
OpenMMLAB系列框架解读(基于PyTorch)
2019杭电多校 第一场 6581-Vacation【思维】
matplotlib.pyplot接口汇总
直接使用CRecordset,不使用CRecordset的派生类
函数之递归[通俗易懂]
Code representation pre training language model learning guide: principles, analysis and code
ATL容器——CAtlMap,CRBMap
Utilisation simple de quartz et de ses emplois
matplotlib绘制sin函数图像
不懂点儿统计学,《星球大战》白看了