当前位置:网站首页>Shortest moving distance and entropy of morphological complex
Shortest moving distance and entropy of morphological complex
2022-07-27 10:58:00 【Black elm】
(A,B)---4*4*2---(1,0)(0,1)

Classify with neural network A and B, Give Way A and B All are 4*4 Matrix , among A In all the grids of 1,B There is 1 individual 0. There are only 1 A picture , Every picture repeats continuously to realize network convergence . In convergence error δ Consistent case , Count the average number of iterations . Get the form
431 | 432 | 433 | 434 | |
δ | The number of iterations n | The number of iterations n | The number of iterations n | The number of iterations n |
5.00E-04 | 49664.99 | 49837.2 | 50115.58 | 49468.1 |
4.00E-04 | 61174.29 | 60516.34 | 60769.59 | 60547.6 |
3.00E-04 | 79569.04 | 78832.1 | 79287.35 | 80434.3 |
2.00E-04 | 116011.4 | 116304.9 | 115951.6 | 115654.2 |
1.00E-04 | 224611.3 | 223524.3 | 225130.4 | 225118.7 |
Put this 4 Draw lines together

this 4 The lines are highly coincident , It reflects the symmetrical relationship .

(A,B)---4*4*2---(1,0)(0,1)
Do the second group again ,A It's the same ,B There are two 0, Statistics 4 In this case , The number of iterations obtained is as follows
421 | 422 | 423 | 424 | |
δ | The number of iterations n | The number of iterations n | The number of iterations n | The number of iterations n |
5.00E-04 | 43419.55 | 43768.71 | 44255.1 | 44394.1 |
4.00E-04 | 53934.45 | 53668.25 | 54321.85 | 54310.12 |
3.00E-04 | 70134.04 | 70128.54 | 70742.93 | 71375.41 |
2.00E-04 | 103634.2 | 104216.6 | 103698.3 | 103876 |
1.00E-04 | 200992.7 | 201399.3 | 201920.7 | 201201.3 |
Due to the existence of symmetry , this 4 The lines are also highly coincident

But compare 43 The data and 42 The data will find
43 | 1.00E-04 | 224611.3 | 223524.3 | 225130.4 | 225118.7 |
42 | 1.00E-04 | 200992.7 | 201399.3 | 201920.7 | 201201.3 |
43 The data of is greater than 42 The data of , Consider the assumption that the number of iterations is inversely proportional to entropy , This shows that 42 Entropy of is greater than 43 The entropy of

For example, classify those involved A and B The numbers in are regarded as components A and B The particles of the universe , And the process of classification is to make A and B The particles in exchange positions with each other , The so-called entropy is the distance of particle motion . For example, for 431, obviously B Medium 3 Particles don't need to move , because A Medium (1,0),(0,1),(1,1)3 The two positions are the same 3 A particle .
And the only thing that needs to be moved is A Medium (0,0) The particle , And the shortest moving method of this particle has two kinds of the same probability , Namely (0,0)→(1,0) and (0,0)→(0,1). The distance of each movement is 1, therefore 431 The corresponding particle moving distance is 1.
And yes 421, Similarly, just move A in (0,1) and (1,1), Two particles . In the same way , The least expensive way to move these two particles is (0,1)→(0,0), and (1,1)→(1,0), Every moving distance is 1, So the total distance is 2, So moving distance and 421>431, therefore 431 The number of iterations of is greater than 421.

The third group of experiments was done in the same way ,B There are only 1 individual 1. Give Way A And B classification
The obtained iteration times table
411 | 412 | 413 | 414 | |
δ | The number of iterations n | The number of iterations n | The number of iterations n | The number of iterations n |
5.00E-04 | 41909.38 | 41784.2 | 42049.96 | 41667.75 |
4.00E-04 | 51967.94 | 52190.18 | 51324.76 | 51861.47 |
3.00E-04 | 67093.26 | 67083.14 | 67088.89 | 66872.62 |
2.00E-04 | 98633.1 | 98606.23 | 101325.4 | 99056.68 |
1.00E-04 | 192559.2 | 193504.3 | 194800.3 | 192974.1 |
Also this 4 The lines are highly coincident

Compare the number of iterations when the convergence error is consistent
43 | 1.00E-04 | 224611.3 | 223524.3 | 225130.4 | 225118.7 |
42 | 1.00E-04 | 200992.7 | 201399.3 | 201920.7 | 201201.3 |
41 | 1.00E-04 | 192559.2 | 193504.3 | 194800.3 | 192974.1 |

obviously 41 The series is the smallest ,41<42<43.

Calculate with the same method that particles move with each other ,A and B classification , Only A Medium 3 Particles need to be directed B The position of the only particle in moves , And the moving distance is 1+1+1.414=3.414. And this value is greater than 42 Series of 2, Greater than 41 Series of 1. therefore 41 The entropy of the series is the largest . The number of iterations is the smallest .
Therefore, the total moving distance 41>42>43, The number of iterations 41<42<43.
边栏推荐
- Learning C language together: structure (2)
- Shardingsphere kernel principle
- YonBuilder赋能创新,用友第四届开发者大赛“金键盘奖”开启竞逐!
- No identifier specified for entity solution
- Document intelligent multimodal pre training model layoutlmv3: both versatility and superiority
- Edata base, a secondary development project based on spark packaging, is introduced
- Ten year structure five year life-07 young and vigorous transformation
- 如何创建一个带诊断工具的.NET镜像
- Deep analysis: what is diffusion model?
- gyp ERR! configure error. gyp ERR! stack Error: gyp failed with exit code: 1
猜你喜欢

Li Hongyi_ Machine learning_ Assignment 4 (detailed explanation)_ HW4 Classify the speakers

One stop monitoring of the software and hardware infrastructure of the whole university, and Suzhou University replaces PostgreSQL with time series database

Beijing publicized the spot check of 8 batches of children's shoes, and qierte was listed as unqualified

Gamer questions

Codeforces Round #807 (Div 2.) AB

JSP自定义标签之自定义分页01

TDengine 商业生态合作伙伴招募开启

Use kaggle to run Li Hongyi's machine learning homework

开源项目丨Taier1.2版本发布,新增工作流、租户绑定简化等多项功能

Views, triggers and stored procedures in MySQL
随机推荐
一起学习C语言:结构体(二)
C语言 2:求三数字最大值,求三数字中间值,编写程序步骤
如何创建一个带诊断工具的.NET镜像
The permission problem of Oracle operating openldap
Analysis of heterogeneous computing technology
Using skills of word
Shardingproxy sub database and table actual combat and comparison of similar products
学习笔记-微信支付
Antd table+checkbox default value display
Have you ever seen this kind of dynamic programming -- the stock problem of state machine dynamic programming (Part 1)
jvm--字节码浅析
ASP.NET Core依赖注入之旅:1.理论概念
搭建 Samba 服务
DNS principle and resolution process
Delete in MySQL: the difference between delete, drop and truncate
神经网络学习笔记
No identifier specified for entity solution
Use__ slots__ And__ dict__ To save space (it's simply a qualitative leap, and leetcode's personal test is effective)
Tdengine helps Siemens' lightweight digital solution simicas simplify data processing process
parsel的使用