User Manual, PG 431, Universitaet Dortmund
Contents
1. Configuration Modes informations file berlin 4 para Modes number Sr Figure 8 Problem configuration dialog After deciding which problem to optimize within this single run ensure that the problem is configured to the needs of the experiment it will be take part in The problem configurator as shown in figure 8 shows all available options with their values for the selected problem To change the value for a specific option simply select the corresponding table cell by clicking it and editing the value If it becomes necessary to reset values clicking the Default button resets all pa rameter values to their factory setting The button is located to the right of the parameter list see figure 5 3 MANAGING SINGLE RUNS 9 3 2 Heuristic In this section the heuristic part of the single run configuration is described All parameters regarding the strategy of the heuristic can be set as described here The single run heuristic dialog is similar to the problem dialog This is where all runtime options of the heuristic can be set Choose Heunstic GA Permutation ES Lreat_Deluge_Algorithm 54_Permutation ACO Ey _ Permutation Lreat_Deluge_Permutation Figure 9 Heuristic selection dialog To the left part of figure 5 a list with all installed heuristic plug ins can be found figure 9 Description Implementation of a GA Genetic Algorithm Heuristic Figure 10 Heuristic description dialog When a heuristic plug2. button in the toolbar figure 20 In case you made a mistake in the configuration of a very long run you can cancel the complete run by pressing the stop button 6066 000 09000 Figure 20 The toolbar 4 MANAGING COMPLETE RUNS 16 4 3 Saving In order to save a complete run choose the Save entry from the file menu see figure 21 or press the disk symbol on the toolbar see figure 20 When saving a complete run for the first time MooN will ask for a file name Otherwise the file on the disk is replaced with the recent complete run Saving an already saved run into a different file is done by selecting Save As from the File menu File Optimization Singlerun Help Figure 21 The menubar After entering a proper name pressing the Ok button of the File Save dialog and MooN writes the complete run to the hard disk 4 4 Loading Complete runs are saved as xml configuration files To load a saved complete run choose Open from the File menu This opens a file chooser similar to the one in figure 2 After selecting the desired file pressing the Ok button will load it Note that loading might not be sucessfull if the file specifies plug ins that are not installed or have different parameters This might be the case if the complete run was created with an installation of MooN that had more or other plug ins installed Installing the missing plug ins will help this problem 5 COMMAND LINE USAGE 17 5 Command Line Usage
3. 2 1 Installing Plug Ins conoser ras PO ww 2 2 Deleting PISTAS 42 24 64 wed mee EHS ERS He ER ee 3 Managing Single Runs uk PODOI i ae eH EEE oe eee oe Se Oe i ka he eee ee eee ee ee ee eS Cee OO SO hh ee eG Of General Options o ssa oe eB KR eK EH ee ee EES 4 Managing Complete Runs Al COMPOSE asa ee oe eee rasos lo PU rr nc oir soso ee a ee ee PA eee eee ee A AAA 5 Command Line Usage 5 1 Invoking MooN from the Command Line 5 2 Un Installing Plug Ins from the Command Line 5 2 1 Installing a Plug In oc o eR eS ee eS 5 2 2 Removing a Plug In un bee ha ed we oe Ewe Ew we Dil Lista FUME e 2 amp ac oe bars sas eR GR eR Rew 1 INTRODUCTION 1 1 Introduction One small step for man one giant leap for mankind Neil Armstrong on landing on the moon Welcome to MooN This is a short user manual which should help you exploring the MooN system MooN is a free software tool that provides a plug in based framework for implement ing heuristics and optimization problems These plugins can be installed into the core system after which they are immediately available to be used for experiments The MooN core system provides all the functionality to run interesting experiments on the installed plugins There are a number of plugins included in the MooN dis tribution as it can be downloaded from http lsii www cs uni dortmund de lehre SoSe03 PG431 MooN moon htm In the problem section
4. Complete run list Adding a single run to the list To add a new single run select Add in the Single run menu or click on the plus button Duplicating single runs A feature which is very handy when composing list of very similar single runs is the copy feature To duplicate an existing single run select it an chose Copy from the Single run menu The clone will appear at the end of the list Changing the order of execution To move a single up and down the list one can use the Move up Move down in the Single run menu or the arrow buttons in the toolbar Removing single runs The minus button or the Delete entry in the Single run menu will remove a selected single run a 0 O 4 MANAGING COMPLETE RUNS 15 4 2 Running After a complete run has been loaded or newly created it can be started by selecting Optimization Run from the menubar Alternatively the play button in the toolbar figure 20 does the same Depending on how complex the complete run is the running process can take a long or short time However the course of optimizations can be followed on the Runtime Visualization Graphic that opens up automatically upon a run as shown in figure 19 value of global best fitness 7 4177906E 6 Number of generations 29 l SingleRun Visualisation Figure 19 Runtime visualization If you want to observe an interesting state of the optimization in more detail the current run can be paused by the pause
5. Value orar Nodes informations file berlinS2 para Nodes number 52 Description Symmetric TSP traveling salesman problem The concret TSP instances are specified in seperate files i SingleRun Figure 5 Single run parameter dialog The single run problem dialog close up shown in figure 5 enables the user to select problem specific options Choose Problem Rastrigin Mold_Temperature_Control tspProblem Pickup_and_Delivery Hevator Permutation_Sort Spherical Michalewicz Rosen brock Figure 6 Problem selection dialog To the left part of figure 5 there is a list with all installed problem plug ins close up in figure 6 If a desired plug in is not found in the list it needs to be installed After installing it see section 2 1 it will be visible in this list and can be used in single runs from there on When a problem is selected from the list a detailed description of the plug in is displayed in the lower left corner shown in figure 7 3 MANAGING SINGLE RUNS 8 Description In the Pickup and Delivery Problem a transportation carrier with one wehicle receives a set of n requests Each request consists of picking up a load with a certain size from ae some origin and to deliver it to a destination in order to respect the time windows associated With the pickup location and the delivery location The aim is ta maximize the number 0 Figure 7 Problem description dialog
6. the description section of a single run one can enter a detailed description of a single run It is recommended to put a single one line short description in the first line since it will be displayed in the complete run overview The length of the description is not limited and the visible are provides enough space for detailed informations regarding a single run figure 16 Enter Description SO run on Elevator Problem PSO Parameters ognitive Terr 2 02 ocial Term 2 02 ertia WYveight 1 2 elocity Bound 1 0 wearm Size 50 Figure 16 Single run description dialog Although descriptions are not necessary for the execution of the single run it is strongly recommended to explain the experiment here in order to be able to under stand the purpose of the configuration without reviewing all values again Repeating Single Runs Another feature designed to conduct real life experiments easily is the ability to re peat a single run several times by changing the repetitions value in the dialog shown in figure 17 This reduces the variance and generates statistically more significant results without creating new single runs with identical parameter settings Copying Single Runs Another handy feature is the Copy item from the Single run menu It is helpful for creating complete runs that consist of similar single runs When applied to a 3 MANAGING SINGLE RUNS 13 Options Repetitions of asing 10 Figure 17 Sin
7. you can find common test functions like e spherical function Schwefel function Griewank function e Travelling Salesperson problem In the heuristic section there are implementations of e genetic algorithms particle swarm optimization e ant colony optimization e evolution strategy and other heuristics You can also find a number of exit conditions such as e number of generations e exit at a given time e fitness threshold MooN has been developed by students of the University of Dortmund Germany It was a project work of a course called Metaheuristics New Ideas for Optimization We wish you a lot of fun with MooN and hope that it makes your scientific work easier 1 INTRODUCTION 2 1 1 Requirements MooN is written in the Java programming language An installation of the Java 2 Platform Standard Edition J2SE version 1 4 2 is needed For details about J2SE downloading and installing instructions refer to http java sun com j2se 1 4 2 index jsp Additionally some third party packages are needed to run MooN Those packages are Log4J Xerces and JDOM Log4J version 1 2 8 Log4J is an Apache Software Foundation project for advanced logging control The main benefit is the logging output can be controlled through a configuration file with out modifying the source code Log4J can be obtained from the project s homepage http logging apache org log4j docs Xerces 2 for Java version 2 6 x Xer
8. Also necessary third party libraries have to be copied to bin lib see below e doc The MooN Documentation user and developers manual readme and the JavaDoc can be found here e misc The jar files of the included plugins and a sample R script for statistical analysis reside here e src The source files of MooN e Uninstaller The uninstaller for automatical removing of MooN Copying the necessary Libraries In order to run MooN some third party libraries must be copied to the moon bin 1ib directory For details about those libraries and where they can be obtained refer to section 1 1 After downloading the described libraries they need to be copied to moon bin 1ib MooN can then be started by either selecting the Moon bat or Moon sh depending on the underlying operating system 2 MANAGING PLUG INS 4 2 Managing Plug Ins Computers are useless They can only give you answers Pablo Picasso One of the main advantages of MooN is the plug in approach Through this approach one can install and remove heuristics problems and exit conditions at any time Given a valid heuristic plug in received through a colleagues email downloaded from the Internet etc installing it into your installation of MooN is a matter of seconds The steps needed are described in detail below 2 1 Installing Plug Ins File Optimization New Open Close ave 5 Save as Install Plugin Delete Plugin Figure 1 I
9. Bad Command or File Name Good try though Anonymous For automation purposes MooN can be invoked from the command line This can be very helpful to conduct complex and time consuming experiments on environments that only allow shell access e g batch systems or mainframes Following operations can be called from the command line e starting complete runs e installing plug ins e uninstalling plug ins e listing plug ins These functionalities are described in detail in the following subsections 5 1 Invoking MooN from the Command Line Starting MooN from the command line is simple All one needs is to specify the path to a valid XML complete run definition This XML file can be created using the GUI as described in the preceding chapters Of course manipulating it with a text editor is also possible see the developers manual for details on the file structure One can call Moon with the parameter r for run Moon r lt pathToCompleteRunXML gt MooN will process the tasks in the XML definition file There will be no runtime visualization only the described output handling is performed Even if no batch system is used one might consider running MooN from the com mand line since avoiding the runtime visualization causes a noticeable performance gain 5 COMMAND LINE USAGE 18 5 2 Un Installing Plug Ins from the Command Line As mentioned above one can install remove and list the plugins 5 2 1 Installing a Plug I
10. MooN User Manual MooN A Framework for Metaheuristical Optimization 2003 04 PG 431 Chair of Systems Analysis University of Dortmund MooN is licensed under the GNU General Public License Preface From the beginning of the development of MooN we had ambitious plans to not only fulfill the requirements of the university course but to provide a tool to the scientific community with practical value There were only a few comparable projects which almed at making the research work on evolutionary algorithms easier and which was what we identified as our main goal in the early stages of the project We wanted to help scientists focus on the real work giving them the possibility to mostly neglect the craftmanship of programming an environment for the algorithms and problems under examination We hope that our concept of the plug in structure lives up to the ambitions we had in mind Anyway we already found the design quite helpful during the last stages of our project We hope that some people will find that as well Projektgruppe 431 The PG 431 is Selcuk Balci Soren Blom Daniel Blum Vedran Divkovic Dirk Hoppe Djamila Lindemann Ulf Schneider Bianca Selzam Thomas Tometzki Marko Tosic Igor Vatolkin and Stefan Walter CONTENTS Contents Preface 1 Introduction 1 1 Requirements 4 lt 6 lt 464 684 64 468646 544 be dee eae La Installing MoN 25462 8s kee BX Rew EERE Eee ES 2 Managing Plug Ins
11. ber_of_ Generations Permutation _Sort Pickup_and_Delivery Preset_Time Psi Rastrigin Eosenbrock SA 54_Permutation heuristic exit condition problern heuristic heuristic exit condition heuristic heuristic heuristic heuristic problern problern problern exit condition problern problern exit condition heuristic problern problem Heuristic Heuristic Hel problern problern Figure 4 Selecting a plug in for deletion 3 Managing Single Runs Run Forrest run Jenny Curran to Forrest Gump The smallest execution unit in MooN is called single run A single run represents one experiment with its parameter options for the respective heuristic problem and exit condition In order to create a single run in MooN one needs to pick at the least one heuristic one problem and one exit condition The single run can be started with its default parameters However if real life experiments are to be conducted the options of the three plug ins will need to be tuned After all options are set correctly the Accept button is enabled 3 MANAGING SINGLE RUNS T 3 1 Problem We start with an explanation of the problem and its parameter configuration Description Problem Heuristic ExitCondition Choose Problem Configuration Rastrigin Mold_Temperature_Control tspProblem Pickup_and_Delivery Hevator Permutation_Sort Spherical Michalewicz Rosenbrock Schwefel Griewank
12. ces is a highly modular XML parser both for XML parsing and generation MooN makes extensive use of Xerces as XML is a main feature for providing ex tensibility and universality Xerces is also an Apache Software Foundation project available on their XML website http xml apache org xerces2 j index html Note that the Xerces2 Java Parser is needed Two libraries out of the binary dis tribution package are of interest xercesImpl jar and xml apis jar JDOM version 1 0 beta 9 JDOM is a complete Java based solution for accessing manipulating and out putting XML data from Java code It is available on the project s homepage http www jdom org l Generally the latest installation of J2SE and the third party packages should work with MooN Since downward compability is not always achieved the exact version numbers are provided as used to develop MooN 1 INTRODUCTION 3 1 2 Installing MooN MooN is distributed as a jar package including an installation routine Typing java jar install jar at the command line will start the installation process Depending on the systems configuration double clicking the install jar file will do the same The graphical installation manager that is started by this is self explanatory Files and Directory Structure Depending on the decisions during installation the following subdirectories will be created in the MooN directory e bin All class files and libraries are placed here
13. gle run repetitions dialog selected single run in the complete run list the copied single run is appended to the list and the desired changes can be applied to it 4 MANAGING COMPLETE RUNS 14 4 Managing Complete Runs There is a theory which states that if ever anybody discovers exactly what the universe is for and why it is here it will instantly disappear and be replaced by something even more bizarre and inexplicable There is another theory which states that this has already happened Douglas Adams In this part managing complete runs is explained here are four main activites with regard to complete runs composing running saving and loading complete runs 4 1 Composing Single runs are administered in a task list called the complete run list figure 18 Every line represents a single run hence the whole list represents the complete run To create a new complete run select New in the File menu or click on the icon for a new complete run in the toolbar complete run list Exit Condition PSO on Rosenbrock Perforrnanc eck Rosenbro Number_of_Generations PSO on valor first rs set Elevator Number_of Caneraliona PSO on Elevator alternative parameters Elevator PSO Number of Generations GD on Elevator first parameter set Elevator Great_Deluge_Algorithm Number_of_Generations GD on Elevator second parameter set Elevator Great_Deluge_Algorithm Number_of_Generations Figure 18
14. in is selected a detailed description of its function is displayed in the lower left corner as shown in figure 10 Similar to the situation in section 3 1 the list shown in figure 11 contains all available options for the selected heuristic Editing values in the list is done in the same way as described there and the Default button has the same effect as well In addition to the dialog elements in the problem section 3 1 there are output handler options The output handler is the part of MooN that writes experiment data to your hard disk for later evaluation Since different heuristics may have 3 MANAGING SINGLE RUNS 10 Configuration Population size 100 solution minimum 10 0 Dimension 10 selection size 10 Maximal mutationincrement 5 1 Probability of crossover Oe Probability of mutation 1 1 solution maximum 10 0 Figure 11 Heuristic configuration dialog different data to examine you can parameterize the output of the single run There are two important settings to consider the output interval and the output file Depending how fine grained the heuristics output should be on a specific parameter the logging interval takes any positive integer including zero A value of zero means that no logging one that parameter is done a positive value n means that logging for this parameter is done every n th generation When considered usefull for later data analysis the output for every parameter can be written into individual files Th
15. is option is enabled by activating the Log categories to separate files switch see figure 12 OutputHan dler Log filename Default log _ Log categories to separate files Default Logging interval Global_best_individual 1 CUrrent_best_individual 1 Figure 12 Heuristic output handler dialog 3 MANAGING SINGLE RUNS 11 3 3 Exit Condition The exit condition is the part of MooN that controls the termination of a single run For example a single run can be stopped after a specific time or after a certain number of generations Exit conditions are also realized as plug ins One exit condition needs to be selected for a single run to be complete Choose Exit Condition Fitness_Threshold Number_of_Generations Preset_Time Chang e_of_Fitness_Threshold Figure 13 Exit condition choose dialog The dialog shown in figure 13 is used to select the exit condition All installed exit condition plug ins are listed When selected a detailed description of the plug in is displayed in the lower left corner as shown in figure 14 Description This exit condition depends on the number of generations Figure 14 Exit condition description dialog As before parameter values of the selected exit condition can be edited in the assocciated list 3 MANAGING SINGLE RUNS 12 Configuration Mumber of generations 100 Figure 15 Exit condition configuration dialog 3 4 General Options Single Run Description In
16. n One needs a valid plug in in form of a jar archive Install the plug in by using the i parameter Moon 1 lt pathToPlugin jar gt The plug in will be extracted and stored in the standard plug in directory of MooN plugins in the Moon directory Installing it into a different directory is done by using the optional d parameter Moon 1 lt pathToPlugin jar gt d lt pluginDirectory gt 5 2 2 Removing a Plug In Removing a plug in requires its full name To obtain a list of all plug ins from the command line see 5 2 3 Knowing the name uninstalling a plug in is done with the u switch Moon u lt pluginName gt 5 2 3 Listing Plug Ins The 1 switch is used for listing all installed plug ins Moon 1 We hope that this manual will be found helpful in introducing the user to MooN and its usage PG 431
17. nstalling a plug in Installing a plug in is as simple as loading a file In the File menu choose Install Plug In figure 1 In the appearing dialog search and choose your plug in jar file and click Open figure 2 The plug in will be stored in the default plug in directory ready for use and should appear in the list during the configuration of a single run see section 3 2 2 MANAGING PLUG INS 5 install plugin A ae suchen im Cas vo A fa E Wil Y ACOjar Y GA jar D Change_of_Fitness_Threshold jar 3 CA Permutation jar 3 Hevator jar 3 Creat_Deluge_Algorithm jar G E jar i D Creai_Deluge_Permutation jar D ES_Permutation jar D Griewank jar D Fitness_Threshold jar 3 Michalewicz jar Dat el name ES jar Dateityp jar ll Maa Figure 2 Selecting a plug in 2 2 Deleting Plug Ins To remove a plug in choose the Delete plug in entry from the File menu figure 3 From the list of installed plug ins choose the one to remove and confirm this by clicking OK figure 4 The plug in files are deleted from the plug in directory File Optimization Install Plugin Delete Plugin Figure 3 Deleting a plug in 3 MANAGING SINGLE RUNS JDelete Plugin Choose plugin to delete ALI Change_of_Fitness_Threshold Elewator ES ES _ Permutation Fitness_ Threshold GA GA_Permutation Great_Deluge_Algorithm Great_Deluge_Permutation Oriewarik Michalewicz Mold_Temperature_Control Mum
Download Pdf Manuals
Related Search