Skip to content

Latest commit

 

History

History
251 lines (195 loc) · 12.4 KB

User-Guide.md

File metadata and controls

251 lines (195 loc) · 12.4 KB
Raw

User Guide

-> In Chinese <-

Run MTO Platform

  • GUI: mto
  • Command line: mto(AlgoCell, ProbCell, Reps, Par_Flag, Results_num, Save_Dec, Save_Name)
  • Example: mto({'MFEA','AT-MFEA'},{'CEC17_MTSO1_CI_HS','CEC17_MTSO2_CI_MS'},30,true,50,false,'MTODataSave')

Add your algorithm

  • Inherit the Algorithm.m class from the Algorithms' folder to implement a new algorithm class and put it in the Algorithms folder or its subfolders
  • Implement the method function data = run().
  • Add labels to the second line. <Multi-task/Many-task/Single-task> <Multi-objective/Single-objective> <None/Competitive/Constrained>
  • Refer to the MFEA or MO-MFEA algorithm implementation

Add your problem

  • Inherit the Problem.m class from the Problem folder to implement a new problem class and put it in the Problem folder or its subfolders
  • Implement the method function Tasks = setTasks()
  • Add labels to the second line. <Multi-task/Many-task/Single-task> <Multi-objective/Single-objective> <None/Competitive/Constrained>
  • Refer to the CEC17_MTSO or MTMO_Instance1 problem implementation

Add your metric

  • Inherit the Metric.m class from the Metric folder to implement a new metric class and put it in the Metric folder or its subfolders
  • Add labels to the second line.
  • Refer to the Obj.m and IGD.m metric implementation

Module

I. Test Module

  1. Algorithm selection
    • Select an algorithm to be displayed in the Algorithm Tree
    • Open the Algorithm, and it will show the algorithm parameter settings. Double-click to modify
  2. Problem selection
    • Select a problem and display it in the Problem Tree
    • Open the Problem Node to display the problem parameter settings. Double-click to modify
  3. Run
    • Click the Start button
  4. Check the figure
    • Task Figure 1D (unified / real)
    • Task Figure 2D (unified / real)
    • Feasible Region 2D
    • Obj Convergence
    • Pareto Front

II. Experiment Module

  1. parameter settings

    • No. of Runs: Number of independent run times
    • No. of Results: Number of save results
    • Parallel：Parallel execution

  2. Algorithm selection

    • After selecting an algorithm in Algorithms, click Add button, it will add the algorithm to Selected Algorithms, you can expand the algorithm and double-click to modify the parameters or algorithm name, double-click to modify the parameters or algorithm name. Multi-selectable, right-click to select all, can be added repeatedly
    • Select the algorithm in Selected Algorithms and click the Delete button to delete the selected algorithm. Multi-selectable, right-click to select all

  3. Problem Selection

    • After selecting the problem in Problems, click Add button, it will add the problem to Selected Problems, you can expand the problem and double-click to modify the parameters or problem name. Multi-selectable, right-click to select all, can be added repeatedly
    • Select the problem in Selected Problems and click the Delete button to delete the selected problem. Multi-selectable, right-click to select all.

  4. Start/Pause/Stop

    • After selecting the algorithm and problem, click the Start button to start running.
    • In the process of running, click the Pause button to pause, and then click the Resume button to continue.
    • In the process of running, click the Stop button to stop running.

  5. Table Statistics

    • Display data with metric
    • Draw Metric Convergence. Select table cell first.
    • Draw Pareto Front Results. Select table cell first.
    • Data type
      • Mean
      • Mean&Std
      • Std
      • Median
      • Best
      • Worst
    • Statistical test
      • None
      • Rank sum test
      • Signed rank test
    • Highlight data
      • None
      • Highlight best
      • Highlight best worst
    • Save the data, click the blue Save button to save the current table content

  6. Read/Save Data

    • To save data, click the Save Data button to save the data of the currently running experiment
    • Read data, click the Load Data button, read the saved experimental data, and display the data

III. Data Process Module

  1. Read data

    • Click the Load Data button, read the saved experimental data, and add it to the Data Tree. Multi-selectable, can be added repeatedly
    • Expand the data in the Data Tree can display the specific content of the data, can modify the name of the data

  2. Delete Data

    • Select the data in Data Tree and click the Delete Data button to delete. Click the Delete Data button to delete. Multi-selectable, right-click to select all

  3. Save Data

    • Select the data in the Data Tree and click the Save Data button to save it. Multi-selectable

  4. Data split

    • Split by the number of independent runs, select more than 1 data in the Data Tree, click the Reps Split button to split the selected data by each independent run and add it to the Data Tree.
    • Split by Algorithms, select more than 1 data item in the Data Tree, click the Algorithms Split button to split the selected data by algorithms and add them to the Data Tree.
    • Split by Problem, select more than 1 data item in the Data Tree, click on the Problems Split button to split the selected data by problem and add it to the Data Tree.

  5. Data Merge

    • Merge by independent runs, select 2 or more data in Data Tree, provided that all settings are the same except the number of runs, click the Reps Merge button to merge the selected data by the number of independent runs and add them to the Data Tree.
    • Merge by Algorithm, select more than 2 data items in Data Tree, provided that all settings are the same except Algorithm, click Algorithms Merge button to merge the selected data by the algorithm and add them to the Data Tree.
    • Merge by Problem, select more than 2 data items in Data Tree, provided all settings are the same except Problem, click the Problems Merge button to merge the selected data by problem and add them to the Data Tree.

中文指南

运行MTO Platform

  • GUI界面: mto
  • 命令行: mto(AlgoCell, ProbCell, Reps, Par_Flag, Results_num, Save_Dec, Save_Name)
  • 示例: mto({'MFEA','AT-MFEA'},{'CEC17_MTSO1_CI_HS','CEC17_MTSO2_CI_MS'},30,true,50,false,'MTODataSave')

加入自己的算法

  • 继承Algorithms文件夹下的Algorithm.m类实现新的算法类，并放入Algorithms文件夹或其子文件夹内
  • 实现 function data = run()
  • 在文件的第2行添加标签 <Multi-task/Many-task/Single-task> <Multi-objective/Single-objective> <None/Competitive/Constrained>
  • 可参考 MFEA、MO-MFEA 算法的实现

加入自己的问题

  • 继承Problem文件夹下的Problem.m类实现新的问题类，并放入Problem文件夹或其子文件夹内
  • 实现 function Tasks = setTasks()
  • 按照Problem类中的各虚函数的注释实现继承的虚函数
  • 在文件的第2行添加标签 <Multi-task/Many-task/Single-task> <Multi-objective/Single-objective> <None/Competitive/Constrained>
  • 可参考 CEC17_MTSO、MTMO_Instance1 问题的实现

加入自己的指标

  • 继承Metric文件夹下的Metric.m类实现新的指标类，并放入Metric文件夹或其子文件夹内
  • 在文件的第2行添加标签 ，对应于列表数据展示和图像数据展示
  • 可参考 Obj.m、IGD.m 的实现

功能

一、测试模块

  1. 算法选择
    • 选取一个算法，显示在Algorithm Tree中
    • 打开Algorithm会显示算法参数设置。双击修改
  2. 问题选择
    • 选取一个问题，显示在Problem Tree中
    • 打开Problem Node会显示问题参数设置。双击修改
  3. 算法运行
    • 点击Start按钮开始运行
  4. 查看图像
    • 问题1维图像（归一化/原始）
    • 问题2维图像（归一化/原始）
    • 可行域2维图像
    • 收敛图，运行完后显示
    • Pareto前沿，运行完后显示

二、实验模块

  1. 参数设置

    • No. of Runs: 独立运行次数
    • No. of Results: 结果保存次数
    • Parallel: 是否开启并行

  2. 算法选择

    • 在Algorithms中选择算法后，点击Add按钮，会将算法添加到Selected Algorithms中，可以展开算法，双击修改参数或算法名称。可多选，右键全选，可重复添加
    • 在Selected Algorithms中选择算法，点击Delete按钮删除所选算法。可多选，右键全选

  3. 问题选择

    • 在Problems中选择问题后，点击Add按钮，会将问题添加到Selected Problems中，可以展开问题，双击修改参数或问题名称。可多选，右键全选，可重复添加
    • 在Selected Problems中选择问题，点击Delete按钮删除所选问题。可多选，右键全选

  4. 开始/暂停/终止

    • 选取算法和问题后，点击Start按钮开始运行
    • 在运行过程中，点击Pause按钮暂停，再点击Resume继续
    • 在运行过程中，点击Stop按钮终止

  5. 表格统计

    • 显示数据，由Metric计算
    • 绘制Metric收敛图，先从表格中选取数据
    • 绘制Pareto前沿结果，先从表格中选取数据
    • 数据类型
      • Mean 平均目标值
      • Mean&Std 平均目标值 (标准差)
      • Std 目标值标准差
      • Median 目标值中位数
      • Best 最优目标值
      • Worst 最差目标值
    • 统计测试
      • None
      • Rank sum test 秩和检验
      • Signed rank test 符号秩检验
    • 高亮数据
      • None 无高亮
      • Highlight best 高亮最优值
      • Highlight best worst 高亮最优值和最差值
    • 保存数据，点击Save按钮，保存当前表格内容

  6. 读取/保存数据

    • 保存数据，点击Save Data按钮，保存当前运行实验的数据
    • 读取数据，点击Load Data按钮，读取保存的实验数据，并显示数据

三、数据处理模块

  1. 读取数据

    • 点击Load Data按钮，读取保存的实验数据，加入Data Tree。可多选，可重复添加
    • 在Data Tree中展开数据可显示数据具体内容，可修改数据名称

  2. 删除数据

    • 选取Data Tree中的数据，点击Delete Data按钮进行删除。可多选，右键全选

  3. 保存数据

    • 选取Data Tree中的数据，点击Save Data按钮进行保存。可多选，右键全选

  4. 数据分割

    • 按独立运行次数分割，在Data Tree中选取1条以上的数据，点击Reps Split按钮，可将选取的数据按照每次独立运行分割，并添加到Data Tree中
    • 按算法分割，在Data Tree中选取1条以上的数据，点击Algorithms Split按钮，可将选取的数据按照算法运行分割，并添加到Data Tree中
    • 按问题分割，在Data Tree中选取1条以上的数据，点击Problems Split按钮，可将选取的数据按照问题分割，并添加到Data Tree中

  5. 数据合并

    • 按独立运行次数合并，在Data Tree中选取2条以上的数据，前提是除运行次数外其他设置相同，点击Reps Merge按钮，可将选取的数据按照独立运行次数合并，并添加到Data Tree中
    • 按算法合，在Data Tree中选取2条以上的数据，前提是除算法外其他设置相同，点击Algorithms Merge按钮，可将选取的数据按照算法合并，并添加到Data Tree中
    • 按问题合并，在Data Tree中选取2条以上的数据，前提是除问题外其他设置相同，点击Problems Merge按钮，可将选取的数据按照问题合并，并添加到Data Tree中