FLAME GUI User Manual
Contents
1. Figure 7 1 Population properties dialog 4 Enter values of global environment constants 5 Create one or more instances of population Each of stages are described in the sections below 7 2 STEP 1 Creating a population First create a new population from file menu When creating a population you ll be asked for the model xml file which must be in FLAME XMML format Then you can give a name to the population and set the number of regions in the window that pops up as shown in Figure The default number o regions is 1 You can increase this number depending on your needs When you are finished you must save your changes using Update and close button in the properties dialog You can later view or change these properties using the Properties button in the toolbar The update operation only updates the population in your session and does not save the changes in the file system To save your changes permanently at any point you must use Save population or Save population as menu items from the File menu group The population information is saved in a file with pop extension in your file system If you want to use a previously created population click the Open population menu in the File menu group then choose a population Alternatively the name of a pop file can be given as a command line argument to PopGUI when invoking the program see section about command line Chapter 7 Po2. 38 Appendix A Manual installation from sources A 1 Installing PopGUI PopGUI is written in Python language for faster development and portability reasons It uses GTK for its graphical user interface Therefore PopGUI will run on any platform for which Python version 2 5 or higher and GTK version 2 0 or higher is available including GNU linux Microsoft Windows most Unixes and others The program itself consists of two files only poplib py which contains the library functions and popgui py which provides the GUI Follow the instructions below for your operating system to start using PopGUL A 1 1 Installation on GNU Linux Install python gtk2 package Then place poplib py and popgui py into some directory and run popgui py If you want to do these from a command line do as follows example for Debian or Ubuntu Linux sudo aptitude install python gtk2 python popgui py A 1 2 Installation on Windows Install the following in order See http www pygtk org downloads html for detailed instructions and links for setting up Python with GTK support 1 Python 2 5 or newer 2 GTK win32 runtime 3 PyCairo 4 PyGObject A 1 Appendix A Manual installation from sources 5 PyGTK After that copy poplib py and popgui py into a directory and run popgui py
3. 2 Edit Regions 3 Edit Constants 4 Edit Memory Variables 5 Instantiate population 0 xml PopGUI version 0 5 0 pared Regions Population home mg NUMBER OF AGENT E 7 a a A s a a A gt a N EACH REGION Region Region 2Region 3Region 4Region SRegion 6Region 7Region BRegion 9Region 10Region 11Region Household IGFirm Eurostat Government Update and close Figure 7 2 Editing population composition Chapter 7 Population Design with PopGUI Agent Variable Memory Variables Var Type Region 1 Region 2 b Firm gt Household id region id neighboring region ids gov id bank id day of month to act payment account mean income week of month weekly budget rationed mall completely sold out order guantity 2 received_quantity 2 day of week to act int SPECIAL int SPECIAL int_array int getSelfvar region_id int getAgentGlobal Bank int double double int double int int ordered_quantity received_quantities int 18 day of month receive income int current productivity employer double current mean specific skills employer double total taxes double 9 9 Syntax help Region copy Expand All Validate Update and close Figure 7 3 Editing memor
4. The validation operation only validates expression syntax and does not attempt a full initialization and checking of expressions and references for fitness Thus you may still catch some problems when you attempt to instantiate population later 7 5 STEP 4 Editing constants The constants defined in your model will be assigned a value of zero initially You can change these values using the Edit Constant toolbar button The constants screen also have a Validate button to ensure the expressions you have entered are valid before you save them Instead of numbers you can also use getAgentCount agent name as a value for constants Unless you set sensible values to constants it is likely that you will get errors like division by zero when you validate memory variable specifivations that use these constants 7 6 STEP 5 Instantiating population and finishing up Once you finish entering memory variables and update your changes using the save and close button you can create an instance of the population i e the 0 xml file in EURACE parlance using Instantiate population button You will be asked for the name of the file to export the population Depending on the number of agents in your population this process can take a long time and the progress will be displayed on the main screen You can cancel this operation using the Cancel button at any time Once you are finished working with your population r
5. gt Consumption_Goods_Market gt Credit_Market Model Version gt Financial Management Dummy gt Financial Market Dummy UNIBI pre Oxford gt Investment_Goods_Market cay gt Labour_Market Modal Dasctiphon gt Government Authors Simon Coakley Mariam Kiran Simon Gemkow gt Statistical_Office Philipp Harting Sander van der Hoog Mario Locci Sabrina v Agents Ecca Marco Raberto Andrea Teglio Saul Desiderio gt amp Firm gt amp Household gt Bj Mall gt amp IGFirm gt amp Eurostat gt Bi Bank gt amp Government gt Bj Central Bank File Path v Messages gt Gi bank account update del tags Prague Demo Model eurace model xml Select Fil gt G central_bank_account_update gt Environment cancel F save _ __ Figure 5 1 XMME AgentGUI Screenshot XMME GUI design is based on two panes Left pane provides a hierarchical tree view widget where all the XMML nodes are represented The right hand pane allows the user to edit available nodes in detail A screenshot is provided below on Table 5 1 Chapter 6 Compiling Agents with GXparser Once a model composed of several agent types and markets is designed it has to be compiled by the so called XParser component of the FLAME The GXparser module of EURACE Toolkit is essentially a graphical user interface wrapper of XParser It eases the manual process of economic des
6. Chapter 2 Introduction This document serves as the definitive documentation for the suite of user level programs that are created as part of the EURACE project effort The EURACE project aims to produce socio economic models and multi agent simulation software for use in economic policy experiments The project has started in September 2006 and will conclude in September 2009 Information on EURACE project can be found at An important requirement for EURACE project was creation of a software platform for design and implementation of agents creation of agent populations creation and execution of experiment sets and evaluation and visualization of simulation results While an existing multi agent framework FLAME Xagents framework was adopted in EURACE project and improved as part of the project effort such tools for pipeline of agent population and experiment design not existed at the time and created as part of the EURACE project The set of user interface tools in EURACE software platform collectively called here as EURACE GUI were distributed individually during early stages of project This documentation provides a sin gle entry introduction to capabilities and usage of EURACE GUI Up to date information about the EURACE GUI can be accessed online at EURACE GUI consists of several independent modules that were created at different stages of the project Various different technologies were employed for each module in order to sati
7. 3 getSelfVar region_id getAgentVar gov_id 7 4 9 Some useful Python constructs Python has a simple syntax and some simple constructs can prove useful in expressing your memory variables We present some constructs here that are observed to be of common use based on feedback from our users For further information you are advised to consult Python tutorial at its website http www python org Constructing lists range start end step constructs a list which includes start but not end If given the list is incremented with the given step value Lambda functions An anonymous function can be constructed using lambda function syntax For example a function to return true if a number is even would be lambda x x Chapter 7 Population Design with PopGUI 25 List processing A new list can be produced by processing numbers in a given list For example to return squares of a range of numbers map lambda x x x range 1 11 If what you want is to select only items from a list that match a criteria use filter For example to select even numbers from a list filter lambda x x 7 4 10 FINISHING UP Validating and saving memory variable specifica tions Before saving memory variables you can use the Validate button on the window which provides ele mentary syntax checking of expressions you enter After validation use Save and close button on the memory variables screen to save your entries
8. Min and Max must be integer and Max must be greater than Min e g deterministic 0 10 lambda x x 10 will initialize the memory variable of a sequence of agents as agent 0 0 agent 1 10 agent 10 100 agent 11 0 7 4 4 Using other memory variables of agent in expressions It is possible to use value of a memory variable of agent itself in specification of another The getSelfVar function is provided for this purpose For example if agent has two memory variables of simple types named numberofchildren and monthlyexpenses one can depend on the other by entering getSelfVar numberof child in the box for monthlyexpenses Chapter 7 Population Design with PopGUI 21 However you must be careful about dependencies of variables PopGUI will analyse which other memory variables a variable depends on and sets their values in proper order But if there are cyclic dependencies which cannot be satisfied you will face an error at some point In the example above getSelfVar returned simply a numeric value since the named variable was a simple variable However if the memory variable referred to is an array or a composite type rather than a simple variable one usually needs to further refer to its elements For example if the returned value is an array getSelfVar somearray 0 will return first zero indexed element of it If the returned value is a composite variable getSelfVar someco
9. a list of tuples variable name condition function Variable name must be a variable of the selected agent and condition is one of the special functions defined These functions can be one of the following equals what checks whether the value of selected variable equals to the value Chapter 7 Population Design with PopGUI 23 between a b checks whether the value of selected variable within the range contains x is the value of selected variable which must be a list contains the value MooreNeighbour numcolumns no checks whether the variable an integer is in the Moore neigh bourhood of no in a geography where regions are laid out in rows with a width of numcomulmns For example if there are 9 regions laid out as follows 1 2 3 4 5 6 7 8 9 then neighbours of region 1 are 2 5 4 whereas neighbours of region 5 arel 2 3 4 6 7 8 9 etc your own functions Other conditions can be defined using so called lambda functions in Python lan guage such as lambda x x lt 100 and x gt 5 For example getAgentRegional Employee conditions skill lambda skill skill lt 100 and skill gt 50 subequals index what checks whether x index what subsubequals index1 index2 what checks whether x index1 index2 what Once an agent is selected its variables can be accessed using a function called getAgentVar variablename For example getAgentGlobal Bank conditions region id Moo
10. a population of agents which will be the initial state of simulation The model created in XMML Editor is used as an input to PopGUI which enables one to specify how different geographical regions in the simulated system are composed of different number of each type of agents how each agent s memory should be initialized and what should be the values of environment constants shared by all agents PopGUI provides a variety of expressions which can be used to specify random distributions for initializing agent memory or to establish relations between agents e g employment relations in an economic simulation Using these specifications PopGUI can create population instances which are different but has the same stochastic properties The model and a population instance is sufficient to run a simulation The model contains what are the memory variables of agents and program code describing their behavior The population instance Chapter 4 Overview of Agent and Population Design Process Figure 4 1 Flow of multi agent design in EURACE XMML Editor Simulation Engine VisGUI Chapter 4 Overview of Agent and Population Design Process 9 on the other hand contains initial values of agent memory for a whole set of agents comprising the population Before a simulation is run however the model must be compiled to generate an executable The GXparser module allows user to do this without going down the elaborate steps of xparser program in FL
11. are not on display when application is closed are going to be deleted permanently Eqrostal average 28 Skill GOUbIe Teporiixt Median 0 0000 0 6142 0 6236 0 6316 0 6400 0 6490 0 8573 0 8650 0 8722 m El bulentex61s Doc 2 visgui tex Docum s screenshots File Br Figure 9 8 Configuring Workspace gt Applications Places System 74 S a Bulent Ozel a ai 1 Mon Oct 19 16 59 8 9 SQLite database file s has been creat nome bulent Code visgui Eurace_Mode Finish Time Mon Oct 19 16 51 25 2009 Available Agent Variables Firm mean_wage double needed capital stock double actual_cap_price double mean specific skills double planned production guantity double production guantity double planned_output double output double unit_costs double total_supply double production_costs double revenue_per_day double technological_frontier double cum_revenue double out_of_stock_costs double malls sales statistics array mall idi malls_sales_statistics array sales arri malls_sales_statistics array sales arr quality double price double price_last_month double demand_capital_stock double planned_production_costs double adaption production volume due to 4 a ee D
12. copies of several types of agents in several geographical regions whose memory and relations with each other are characterized by statistical features derived from actual data The PopGUI component enables simulation designers to describe composition of populations and statistical features of agents and produce population instances that are used as initial states Chapter 1 Executive Summary 2 for simulations ExpGUI Policy experiments require comparison of outcomes of alternative simulations where the com pared cases correspond to covariation of several parameters among their corresponding ranges ExpGUI facilitates creation of a series of simulation tasks from description of the experiment thus reducing errors in creation of policy experiments VisGUI The VisGUI component provides a tool for visualization and analysis of simulation outcomes ConGUI Since simulation development requires several iterations over the various tasks described above the ConGUI component was added to the EURACE GUI to ease launching of the various compo nents described above Figure 1 1 EURACE GUI Figureli iJis the flow diagram and the graphical user interface of the toolkits developed for EURACE The deliverable presented here first describes the architecture of EURACE GUI in relation to other com ponents of the EURACE Software Platform then proceeds to instructions for using various components of the GUI as part of simulation development work
13. functions of the EURACE software platform Design of the EURACE GUI reflects our intention that these software tools reflect most general definition of the particular problems they intend to solve in agent based approach to the extend that it is possible while at the same time they are complete in the sense that they can be deployed in the EURACE project In particular that they work with the FLAME simulation framework Following are the components of EURACE GUI that correspond to different aspects of simulation development work XMML Editor Agent design requires a tedious process of describing agent memory and behavior in an XML format specific to FLAME the so called XMML format while controlling for consistency of design in terms of parameters that affect the overall model The XMML Editor provides a graphical interface which allows the agent designer to focus in his her job in addition to providing facilities for detecting errors in the design It was named after the fact that it translates the design into the XMML format which is usable in FLAME framework GXparser Once a model composed of several agent types is designed it has to be compiled by the so called XParser component of the FLAME The GXparser provides a graphical tool to ease this manual process and report possible compilation errors back to the designer PopGUI A population corresponding to the European Economy in EURACE simulations is composed of multiple thousands
14. producing IATpXtables for constants population composition and memory variables in the population specification When the program is run without any parameters it will ask a few questions to produce desired output When given a tex file as command line argument output is written to that file instead od screen 7 8 NOTES 1 Avoiding dependencies whenever possible The group of functions get AgentGlobal Regional getAllAgentsGlobal Regional create dependency between agents We have seen modelers using these functions just to retrieve id s of other agents However the correct way to do that is to use getAgentIDListGlobal Regional This is because all agents are created and given id s before any of them are populated with memory variables Therefore retrieving agent id s does not generate any dependency and is a faster operation 2 Resolving dependencies If you end up with a cyclic dependency the first thing you can try is to use delayedExecution 3 Be careful with Lambda functions When describing conditions in agent selection you may resort to using Python Lambda functions However you must beware that lambda functions are evaluated when they are called not when they are created For example if you want to choose agents whose regionid is same with the choosing agent you d try a function like lambda Chapter 7 Population Design with PopGUI 27 x x regionid getSelfVar regionid However the function is execu
15. same order get AgentIDListGlobal agentname agentname Get a list of agent IDs You may specify name of one or more agents Please note that agent names are case sensitive i e if you named an agent as Firm you must specify the exact name The function is deterministic ie when called several times the list of agent Ds will be in the same order Since the last two functions return a list rather than a value they are usually intended for array initial ization 7 4 7 Accessing other agents and initializing agent relations Agents are not isolated For example employees and firms in an economics simulation are related to one another through employment relations To address this aspect of populations PopGUI provides a means to select other agents in the population using some criteria and use their memory variables in specifying another agent s memory In order to pick an agent one can use get AgentRegional agentname conditions varname condition get AgentGlobal agentname conditions The first one selects an agent from the same region with the referring agent and the second selects one from the whole population The selectios are random Both functions can be provided with zero or more conditions on the agent to be selected using the following format getAgentRegional Bank conditions givescredit equals 1 badreputation equals 0 I oe e Please note that conditions is
16. AME Economic simulations are useful when one can compare outcomes of different environments popula tions or behavior The ExpGUI program provides facilities to create a set of different initial conditions populations and to run simulations easily Finally the VisGUI program can be used to visualize and analyze the results of simulations The program use plotting and statistical facilities of R statistics package Chapter 5 Agent Design with XMML Editor XMME X Machines Modelling Editor provides the functionality of designing and implementing agent specification and behavior It is a GUI application and the XMML X Machines Markup Language provides the underlying modeling schema In particular XMME helps its user to visually edit an XMML formatted file XMML is a highly verbose and partially untyped data specification language based on XML format This sort of nature makes it relatively difficult to edit individual XML files to describe agents specification and behavior which are very sophisticated in so far that the agents have internal variable and external message depen dencies Furthermore the nested model fascility of XMML specification makes it virtually unmanageable if it comes to large models The XMME wraps the XMML specification and functionality by means of nested data models In its runtime all the decoupled XMML files and XMML nodes are managed by the application so that the model specification remains consistent ov
17. FLAME GUI User Manual Biilent Ozel Mehmet Gencer Vehbi Sinan Tunalioglu and Kaan Erkan TUBITAK UEKAE The Turkish Academic Network and Information Centre Kocaeli Turkey Contents List of Figures 1 Executive Summary 2 Introduction ST ee 4 Overview of Agent and Population Design Process Agent Design with XMML Editor Compiling Agents with GXparser CA Using PopGuly ac gaa a ce ne he he eee bd dd se Ae ERR Bow x RERE TE T EEEE a E eee RA aa e e a A ializati 7 4 4 nz other memory variables of agent in expressions AAA A 7 4 6 Using model constants or population size for scalability purposes 7 4 8 A note about managing dependencies o o e 0004 7 4 9 Some useful Python constructs 7 4 10 FINISHING UP Validating and saving memory variable specifications 7 6 STEP 5 Instantiating population and finishing up 7 7 Command line options utilities anddebuggingl E8 NOTES eiii a e BA ta PS Pea ee A e de REE Ba ae Eales 8 Running Simulations through ExpGUI 9 Analysis of results using VisGUI A a ak A A Se its A 9 2 AA A RN 93 Using VisGUl 2 sake dede bd ee a ESS See See a Ors Platiorms 4 5 5 4 ee a ee ok ee A eee oe eer eee eos A w aD ow 10 12 13 14 15 16 16 19 19 20 20 21 21 22 24 24 25 25 25 26 26 28 CONTENTS 9 3 2 External Packages 2 1 o 9 3 3 Standa
18. OP file has been selected Figure 8 1 Experiment Manager ExpGUI inherits an initial population file from PopGUI It principally facilitates designer to run the designed model for desired number of iterations within a policy experimentation Policy maker is able to test the designed model on following conditions 28 Chapter 8 Running Simulations through ExpGUI e sensitivity to policy parameters e sensitivity to initial population e sensitivity to stochastic process through out iterations Figure 8 1 is a snapshot of the experimet manager under Windows 29 Chapter 9 Analysis of results using VisGUI 9 1 Overview The VisGUI design takes an advanced GUI workspace approach where designers or policy makers can import visualize analyze edit and export simulation results The VisGUI architecture is designed aiming an overall modular implementation It employs a Model View Controller MVC scheme which enables its modularity and extendability Although the package is a sub module within EURACE Software Toolkit it can be installed and run independently This design decision is taken for several reasons First this flexibility is targeted to allow it to be integrated for other multi agent computational tools as an analyses and visualization tool For that reason importation and exportation of standard data formats are adopted for better interoperability Second it allows designers or policy makers easily install and run VisGUI
19. an be downloaded from http www mingw org On Ubuntu or Debian Linux platforms you will need to have the package named build essentals installed Once you have this built environment for your platform follow the instructions in Appendix A Chapter 4 Overview of Agent and Population Design Process Certain stages of multi agent simulation design envisioned in EURACE depend on output of other stages Figure 4 1 visualizes the flow of work through modules in EURACE GUIL also indicating the type of data that comes out from modules and used as input to others This visualization is exactly what you will see when you launch the ConGUI which acts as the control panel for invocation of different modules in EURACE GUI The first stage in EURACE multi agent simulation design is creation of agents and shared environment using XMML Editor Agent design involves 1 designing agent memory for different types of agents 2 describing agent behavior in the case of EURACE FLAME the language used for such description is C programming language using a few FLAME specific constructs 3 designing global environment variables shared by agents 4 designing data types that occur in memory variables and exchanged as data units between communicating agents Use of XMML Editor to carry out these tasks is described in Chapter 5 Once the multi agent model design is completed the model is saved in a FLAME specific format called XMML The next stage is designing
20. aphic features are different since each country have a different population number of people number and average size of firms etc To account for such realistic populations PopGUI allows its user to create several regions and specify numbers of each type of agent separately for each region therefore allowing each region to have a different composition The second phase of specifying a population involves assigning values to agents memory variables Each type of agent is described by a different set of variables For example employees can be described with their skill set income and employment status whereas a firm is described by number of workers production capacity etc For a realistic population one usually needs standard or empirical random distributions For instance in our economic world example sizes of firms in a country usually shows a 13 Chapter 7 Population Design with PopGUI 14 normal distribution although its mean is different for each country region PopGUI provides a rich set of expressions to specify memory variables Another feature of a realistic population is that the agents in it are not isolated but relate to one another For example workers are employed by firms in their region or firms borrow money from banks in their country Therefore these relations must also be initialized in order to create a population and they must be stored in agent memory variables In the case of employment relation example firms wi
21. ation instances created by the program are exported only in the XML format adhering to FLAME specifications Although agent models contain behavior in addition to memory of agents only the agent memory is of interest when creating populations Agent behavior is used later by the simulation engine along with the population created by PopGUI Working out a population with PopGUI consists of several stages 1 Create a population by choosing a model choosing the number of regions in simulation and optionally importing previous work from an existing population 2 Specify population composition by entering how many agents of each type will exist in each region 3 Specify how agent memory variables will be initialized This also means specification of relations between agents since relations are expressed as an agent s identification number appearing as a value in some other agent s memory Chapter 7 Population Design with PopGUI 15 ai L PopGUI v0 5 0 Mel File Help Model Summary 1 Properties 2 Edit Regions 3 Edit Constants 4 Edit Memory Variables 5 Instantiate population 0 xml PopGUl version 0 5 0 started Population properties Population home mgencer test pop is read successfully Name of population European economy Number of regions 27 Model file media disk mgencer EURACE SVNeurad D a import aik Update and close cancel
22. ay in addition to the variable which is used for each element of the array Special features are provided when one wishes to address all elements of an array such as when they will be selected from a set without replacement Most common use of STATIC or CONSTANT variables occur in array sizes in cases when array size is fixed in the model either hard coding or by tying to a constant 7 4 2 Specification syntax for variables and basic random distributions The entries for memory variables are arithmetic expression which produce a single numeric value except the special case of array initialization Therefore the expressions entered into the boxes in this screen can be as simple as a constant number or a simple arithmetic expression e g 2 or 2 2 3 14 4 Whenever necessary PopGUI will convert integer values to real numbers or vice verse In most cases the values will not be constants Use of random or deterministic initializations are very common Following functions are provided for use in such cases in initforms rand int int a random integer from the inclusive range rand real real a random real number from the range choice v1 v2 Pick a value from the list randomly normal mu sigma Normal distribution mu is the mean and sigma is the standard deviation Chapter 7 Population Design with PopGUI 20 discrete probability value probability value Choose value given discrete probabilitie
23. been creati home bulent Code visgui Eurace_Mode i wists yay Finish Time Mon Oct 19 16 51 25 2009 Available Agent Variables Firm mean_wage double needed capital stock double actual cap price double Name office laserjet Properties mean_specific_skills double planned_production_quantity double SILST Gute production_quantity double Type HP HP LaserJet P2015 Series planned_output double output double Output file home bulent Code visgui trunk print pdf si unit costs double total supply double revenue per day double technological frontier double cum_revenue double out_of_stock_costs double malls_sales_statistics array mall_id i malls_sales_statistics array sales arr malls_sales_statistics array sales arr guality double price double 0 200 400 600 800 1000 price_last_month double demand capital stock double iterations planned_production_costs double adaption_production_volume_due_to Frequency value 0 5 0 2 1 0 1 0 0 5 VU Ur ns Figure 9 7 File Operations Chapter 9 Analysis of results using VisGUI File Tools Settings Visualisation Analysis Window Help Vere BUA EBA amp Recycle Session Files Pick files by type y Remove session plots iv Remove session reports The files which are newly created during current session and
24. ed a ranean eae previously picked working foldet you want to replace nome bulent Code visgui exper No to All No Yes to All Yes Previous simulation location is a home bulentCode visgui Eurad Browse iterations O xml 7469k z 100 xml 55k a E VisGul Create Experiment Databas o x 1000 xml 55k Sc Sen a eae eae aes gt 1020 xml 55k Experiment 1040 xml 55k home bulent Code visgui experiment 1 1060 xmI 55k 1080 xml 55k home bulentiCode visgui experimentipop2 par3 45 xml is do men ise Time Left estimated 00 46 01 1120 xml 55k 1140 xml 55k Create Experiment Database 1160 xml 55k 1160 xml 55k 120 xml 55k 1220 xml 55k Status Gathering experiment summary 1222 oe 6 as Time Left estimate hh mm ss 1240 xml 55k 1260 xml 55k 1280 xml 55k 1300 xml 55k meds 2png PIO Xx dom D m 88 bulentex61s c 2 visguitex Docu es KnowledgeDiffusion EJ Xpdf D8 2 pdf LE eurace Data vis E Figure 9 1 Parallel Database Creation retrieves information on the set of agents and their available memory variables This automatically detected set of memory variables are used automatically to populate variable menus Database Creation The creation of a database is necessary when there is a very long and large simulation It is also necessary when there are a set of policy experiment results Created database speeds up time to visualize and a
25. emember to save it before you quit the program Later you can re open them from the file menu and create new population instances for the same population Different instances will not be the same but will have same statistical distributions adhering to your specifications Chapter 7 Population Design with PopGUI 26 7 7 Command line options utilities and debugging PopGUI accepts a few command line options when run from command line instead of from within ConGUI Use python popgui py h to display help on these options Option v will display program version and d will turn on debugging so that program will dump a lot of messages in the terminal it is being run which can be helpful in reporting bugs Also you can give a population file as command line argument for the program to open the population during start up For example python popgui py d test pop will open the population saved in test pop and will turn on debugging while you work PopGUI uses versioning for the saved populations to identify the data structure changes that are not backwards compatible and denies to open populations that were created using past versions of the program and are not compatible with the current version You can disable version checking using i parameter at the command line however it is strongly recommended that such usage should be avoided Distributed with the PopGUI is a small program named popdoc py which is for
26. er time By doing this XMME does not limit the XMML capabilities which allow to represent sophisticated agent models on the contrary it helps modeler to deal with the technical details and internal model specification consistency XMME is a free and open source software application which is developed using C and Trolltech QT development framework The development idea is based on content presentation seperation ie the model internals are controlled by a seperate routine which is a software library called 1ibxmm2 1ibxmm2 provides the model whereas XMME presents it to the user At the same time it passes all the user input to the libxmm2 libxmm2 is written purely in C and core QT library which means that it can be used by any other software routine such as web server applications command line programs etc During the EURACE software toolkit development phase developers have made extensive use of this library with fast prototyping principles Beyond 1ibxmm2 XMME uses QT s model view classes which makes it easy to align to XMML while new functionalities added to XMML specification Both XMME and and 1ibxmm2 are cross platform applications So far it has been ported to Microsoft Windows GNU Linux and MacOSX operating systems 10 Chapter 5 Agent Design with XMML Editor 11 AgentGUI X Machines Model Hierarchy Edit Properties xMML Information Data v Y Eurace Model Name v Nested Models Eurace
27. ew SQLite database file s creation succeeded for the experiment data at Browse Database creation is complete Cancel the plot on nome bulent Code visgui e An histogram of Eurostat no_househol Available Agent Variables Firm mean_wage double needed capital stock double X 4 Di 400 600 iterations a bulent X61s Doc visguitex Docum screenshots File Br BB EURACE Data Visua BM Figure 9 2 Parallel Database Creation desired iteration See Figure 9 4 e Advanced Visualization of Policy Experiments The application also allows designer to examine impact of a policy parameter on macro results sensitivity of the computational model to an initialization and effects of random interactions in between agents across different runs by advanced visualization tool of the application The feature allows to examine multiple variables at the same time very quickly See Figure 9 5 e Automatic Report Generation of Data Points This feature allows designers to view sum mary statistics and exact data points of visualization of a variable e Exporting Raw Data of Plots The feature allows to export data points of a plot in standard data format presumably to be tested or re visualized by a different application See Figure 9 6 e Managing Generated Reports and Plots Analyses and visualizations of an exper
28. ign and report possible compilation errors back to the designer See Figure Model File Path C FP6 EURACE Dissemination FETO3 Demo Eurace_Model_Demo eurace_model xml e Run GNU Make XParser Output Messages Directory Layout Compilation Output ERRORS E Name Type Date Modified Bank agent_headerh hFile 25 10 2009 16 40 54 Central Bank agent header h h File 25 10 2009 16 40 54 Ji Cons Goods UNIBI File Folder 20 04 2009 05 20 00 Ji Credit Ancona File Folder 20 04 2009 05 20 00 OUTPUT Doxyfile File 25 10 2009 16 40 53 Eurostat_agent_header h h File 25 10 2009 16 40 54 pi plaka sapi Financial Management GREQAM_dummy File Folder e ei Financial UNIBI File Folder Imput lle eurace models Firm_agent_headerh h File directory Government GREQAM File Folder templates C Program Files FLAME xparser Mn kene e Loess a RT Household_agent_headerh h File Mdf negentYmodal murace Input model file Cons Goods UNIBI model xml enabled 1GFirm agent headerh h File Input model file Credit_Ancona model xml enabled Inv_Goods_UNIBI File Folder Input model file Financial Management GRE AM dummy model xml enabled Labour_UNIBI File Folder Input model file FINANCIAL UG model xml disabled Makefile 2KB File Input model file Financial UNIBI model xml enabled T 4 1KB Input model file Inv Goods UNIBI model xml enabled Mall agent headech ery Input model file Labour_UNIBI model xml enabled Statistical_Office_UNIBI File Folder Input model file Go
29. iment results in many reports and plots This set of feature allows user either visualize or close them easily or automatically store them on a secondary memory space e Automatic Storage and Re storage of State of Application This feature allows both the main application and all interactive GUI dialogs to store its latest state for subsequent uses Chapter 9 Analysis of results using VisGUI 33 3 Applications Places System FO ide 2 A EURACE Data Visualisation and Analysis Workspace ACE EX B lent ozel JE wil i Fri Feb 27 22 47 File Tools Settings Visualisation Analysis Window Help Y VOQaaHs 4 7 18 Firm planned_output report _ 0 L09 Firm planned output png P d Min 1stQ Median 3rdQ Max e 63 Household general skill png yar Firm planned outp xml 1 xml 7618k About Viscur x 1st 3rd Qu w Median amp VisGUI EURACE Data Visualisation and Analysis Tool v 0 4 0 ia Outer Hinges 250 GNU GPL This application is designed to visualise to analyse to export results of EURACE simulation outputs Frequency 100 Python 2 5 2 Qt 4 3 4 PyQt 4 3 3 R 2 7 1 rpy 2 0 on Linux value l A w 21 xml 7636k 22 xml 7635k n 23 xml 7634k 24 xml 7633k 25 xml 7632k z aa 26 xml 7632k 27 xml 7632k T T T 28 xml 7632k j a 2 i T T T T 29 xml 7632k
30. iterations 1 2 3 4 3 xml 7620k 30 xml 7632k general_skill DA some ROA gt 4 La tbulent xsis Co _ EURACE Data Visua Ci Figure 9 3 Visualization of Time Series a Variable e GUI Interaction and Validity Checks All of GUI dialogs are interactive to user preferences and automatically checks validity of user actions e Writing and Editing Reports VisGUI Workspace provides a text editor with basic functions to enable user to write reports or edit automatically generated reports for convenience e Logging Status and Actions in the Workspace All major user actions are logged and displayed The user can scroll and examine his her previous actions e File Operations All basic file operations including printing options are provided See Figure 9 7 e Configuring Workspace The application provides configuration options for the workspace its state and automatic management of produced plots and reports See Figure e Help Menu VisGUI provides an HTML based navigable help interface 9 3 Using VisGUI In this section we will very briefly demonstrate how VisGUI can be used Chapter 9 Analysis of results using VisGUI 34 gt Applications Places system 7 Q B ao X Bulent Ozel F y mil H Mon Oct 19 16 39 m EURACE Data Visualisation and Analysis Workspace Acc File Tools Settings Visualisation Analysis Window Help Baal VOaa9Hs 4 Log Euros
31. ll have employees whose skillset is suitable for what firm produces and furthermore the relation is exclusive i e if a person is employed in a firm he she cannot be employed by another In the firm bank relation example firms usually borrow money from banks in their own region and the relation is not exclusive i e other firms also borrow money from the same bank PopGUI provides a rich set of features for creating relations between agents both exclusive and otherwise In additional to those substantial features mentioned above some rather practical features were also implemented to address needs of agent modelers For example agent models change through time New agent types may be added or agent memory variables are changed It would be quite impractical to start over with the population design when such changes occur For this reason PopGUI allows importing population specifications from populations that were worked out for older versions of models With the help of this feature populations can be re used with only incremental changes Such features are introduced in the relevant sections below 7 1 Using PopGUI In order to create populations with PopGUI you need a model describing your agents their memory and behavior Since PopGUI was written as a part of EURACE project it currently can use only models given in XMML format which is the format of the FLAME framework the official simulation environment of EURACE Similarly the popul
32. lone Installation E 1 A kema gm kada mM gd ee oO ee ee e A Manual installation rom sources Al stalin PopGUll c resan eiua eeg ke me AA aoa Soe ee Rese A A 1 1 Installation on GNU Linux A 12 Installation on Windows iii List of Figures LI EURACE GUI cs eee eee eee eee eee EEA a A ee e A 2 oat Rica oan deed ae O en as 8 5 1 XMME AgentGUI Screenshot 2 0 000000000002 a 11 01 GUltor AParser ss oA Se Ge ee EE a As 12 7 1 Populationpropertiesdialogi E 15 E E NN 17 Mi a II ti O a 18 parda oA dado we ara cda a 28 wae eae Ge e Be eee ca de ett eae das A 31 biG Gi BG ode eee SSE Rb wee e ar a Base dae 32 ba eats e E es ee eee Te 33 Ld tector bo dc ele et dde e de as da idas amp da Y 34 9 5 Visualization of Advanced Time Series Plots of an Experiment 35 Peek eae Peed do ee Lae 36 a ex sa a eo ve 37 last edad irradia dd oe 38 lv Chapter 1 Executive Summary Creation of software utilities for use in design of agents design of agent populations preparation of simulations and visualization of simulation results has been an essential objective of EURACE project in order to present a complete platform to facilitate use of agent based simulations in policy design The EURACE Graphical User Interface GUI presented here refers to a collection of compatible software tools which provide the components that correspond to these
33. mat and provides facilities to specify composition of population and specifications for initializing each agent s memory variables Populations created by the program is then exported as an XML file so called 0 xml file in FLAME which can be fed into a simulation engine which uses these initial conditions and agent behavior described in models to carry out the simulations When reading through this guide it is important to distinguish the terms population and popu lation instance PopGUI s central concern is to maintain specifications for creating initial populations Population refers to these specifications and the program stores them in its own format in pop files Many population instance s 0 xml files can be created from the same population specification Although PopGUI is designed to create valid population instances the program does not read them back or allow their manipulation Population specification involves two main phases First is population composition In the FLAME framework an agent population can be divided into any number of sub populations For example the EURACE model contains geographical regions such as countries and one can distribute the agents among these regions using the PopGUI One most straightforward example is countries of the world Although each country s social life is composed of same types of agents such as workers consumers firms banks etc their compositions and demogr
34. model population and experiment design process before proceeding to subsequent chapters which explain usage of individual modules Chapter 3 Requirements and Installation EURACE GUI is written with portability in mind Despite the variety of libraries it depends upon all such dependencies are available for common computing platforms including 32 bit or 64 bit variants of GNU Linux Unix or BSD and Microsoft Windows While it is likely that it can be run on other systems such as Mac OS it is neither tested on those platforms not we provide installation packages for them Manual installation of individual modules of EURACE GUI and libraries required by them can be cumbersome Nevertheless details of doing so are described in Appendix A and module sources can be downloaded from http eurace cs bilgi edu tr Before installing EURACE GUI using installation packages you are recommended to install the following software separately FLAME FLAME multi agent platform is developed and distributed by University of Sheffield Please check http ccpforge cse rl ac uk for FLAME installers R R statistics package installers are available for a variety of platforms and is distributed under the GPL license R installers can be downloaded from http www r project org Although you can install and run EURACE GUI without having the above software you should be aware that created simulations cannot be run without FLAME platform and most feat
35. mposite x will return its sub field named x and so on The getSelfVar cannot be used to refer to the variable itself for example to access sibling variables in a data structure etc For this purpose another function is provided getSibling sibling var name When used in an array of data structure this function will retreive the data field of the current array element In cases where a data structure contains other data structures or elements getSibling can instead access higher levels in the data structure hierarchy using level parameter to climb up the hierarchy For example getSibling uncle X level 1 and getSibling grand uncle Y level 2 will seek entities at the level of parent and grandparent in the data structure hierarchy respectively 7 4 5 Initializing arrays The only case where the result of your expression is allowed not to be a single numeric value is when you want elements of an array specified at once In this case your expressions must yield a list whose length is the same with the array length For example if you have an array of length four you can use something like 1 2 3 4 to initialize the array elements sequentially from this list A few functions whose return values are lists are provided for possible use in such cases sample 1 2 3 k Return a k length list of unique elements chosen from the sequence Used for random sampling without replaceme
36. nalyze the results It also allows portability of experiment results in a standard and light format to other platforms and visualization applications We have taken a novel approach to integrate database creation into the application It exploits multi threading and parallelisation paradigms both to decrease the time of database creation and to keep doing analysis at the main workspace during any lengthy process of database creation See Figure and Figure 9 2 Visualization of Time Series of Memory Variables The nature of simulation necessitates a quick trace of some variables over iterations before further examination or more advanced policy experiments User can directly check raw simulation data to examine time series of a memory variable See Figure 9 3 Visualization of Distribution of Memory Variables In the same manner as a quick time series visualization of raw data users are also allowed to examine distribution of a variable at a Chapter 9 Analysis of results using VisGUI 32 a Bulent Ozel F y dil 4 Mon Oct 19 16 52 EURACE Data Visualisation and Analysis Workspace LJ File Tools Settings Visualisation Analysis Window Help 4800990 4 Log E An SOLite database is being created for home bulent Code visgui experiment art Time Mon Oct VisGUI Create Experiment Database VisGUI Into home bulent Code visgui Eurace_Model_Demo_runs experiment 1 Q 9 n
37. nguage More specifically the software is implemented using Python 2 5 with the help of following packages e GUI PyQt4 QtGui e XML PyQt4 QtXml e Statistics RPy 2 DB SqlLite Chapter 9 Analysis of results using VisGUI 35 Applications Places System ER ga ao a Bulent Ozel fl y il Mon Oct 19 16 47 g EURACE Data Visualisation and Analysis Workspace Nel File Tools Settings Visualisation Analysis Window Help Vi BUDGBBA S amp S 4 Log An SQLite database is being created for home bulent Code visgui experiment meds Zipng VisGUI Advanced Time Series and Ensemble Plots Pick an experiment Start Iteration Multiple Variable Plots 2 _Demo_runs experiment Browse 2 End Iteration 241 2 1st Quartile Q1 Pick an initial population H e Frequency 20 Sil Means a 1 i Medians 0 8 Pick a parameter 3rd Quartile 03 1 Selected Variables 0 6 Pick a batch run Household bank idiint Single Variabile Plots ll d Government gdp double Variables Government gdp_forecast double Box Whisker Iterations F Government hh_subsidy_payment double D Agent Memory a Box Whisker Batches gdp double Box Whisker Parameters gdp_forecast double R gdp fraction consumption doub Box Whisker Populations gdp_fraction_investment double gdp_gr
38. nt permutation x y Z Return the same list with elements randomly re ordered sequence start end increment realsequence start end increment generate a sequence of in tegers with given increment within the inclusive interval start start increment start 2 increment END where END lt end rsequence is the version which works for real numbers 7 4 6 Using model constants or population size for scalability purposes You can access model constants or number of agents in the population using the following getConstant constantname Returns the value of model constant e g getConstant alpha to retrieve the value of constant named alpha These constant values are what you set in the Chapter 7 Population Design with PopGUI 22 Edit constants tab of the program get AgentCountGlobal agentname Get global count of agents with given name i e sum of number of agents in all regions e g getAgentCount Firm getNumRegions Returns the number of regions in the population get AgentCountRegional agentname Get regional count of agents with given name get AgentIDListRegional agentname agentname Get a list of agent IDs You may specify name of one or more agents Please note that agent names are case sensitive i e if you named an agent as Firm you must specify the exact name The function is deterministic ie when called several times the list of agent Ds will be in the
39. owth double gov_interest_rate double value 0 4 gt 0 2 a D hh transfer payment double sa GM bulentex61s Doc 2 visguitex Docum screenshots File Br J EURACE Data Visua CHER Figure 9 5 Visualization of Advanced Time Series Plots of an Experiment 9 3 3 Standalone Installation The modular design approach of VisGUI allows it to be installed and used as a standalone application This flexibility will allow the package to be later developed as an platform and economic model indepen dent workspace which can be used to analyse and and visualize results of other computational economics platforms 9 4 Summary The section has presented a brief overview of VisGUI The modular approach of the its implementation and its major design principles will provide us to further develop it and share it with computational economics community in the future Chapter 9 Analysis of results using VisGUI 5 Applications Places System Bulent Ozel 5 4 e il wt Mon oct 19 16 57 U B EURACE Data Visualisation and Analysis Workspace 54 Wa iner gedos File Tools Settings Visualisation Analysis Window Help a voidad He Edrostat average Wage png 7 Log Eurostat average wage 9 SQLite database file s has been creat home bulentCodevisg
40. process specific variables of the agents selected you will need Pythonic expressions as in the following example Chapter 7 Population Design with PopGUI 24 sum agent getAgentVar somevar for agent in getAllAgentsRegional Bank conditions givescredit equals 1 badreputation equals 0 1 The above expression is based on a Python construct which computes a list from elements of a given list e g ixi for i in 1 2 3 7 4 8 A note about managing dependencies The group of functions getAgentGlobal Regional and getAllAgentsGlobalRegional create a depen dency between agents In some cases this creates cyclic dependencies which cannot be resolved by the PopGUI If desired you can solve such situations by using one of the two functions First is signaling a delayed execution of memory variable initialization delayedExecution some expression the expression is executed after all agents and their variab to evoid syntax errors due to quotes in your own expressions use long string syntax in Python by putting your expression in triple quotes e g delayedExecution getAgentGlobal Bank getAgentVar id Another function is available to fine tune dependencies at the level of variables instead of agents dependencyFineTune type toagent tomemvar expression e g dependencyFineTune Global Government gov id X getAgentGlobal Government conditions regions MooreNeighbour
41. pulation Design with PopGUI 16 options below Once you have a population loaded to PopGUI pressing Model summary button will display a rudimentary structure of the model you have chosen for inspection purposes Note that there s also an import button on the properties dialog The import feature is provided for a practical concern in modeling work Once a population is created PopGUI forgets about the original model and only remembers agents and their memory variables However the agent models may change while work on population continues It would be quite impractical to start over entering population specifications every time the model changes and when the changes are minor e g a few memory variables are added or updated in the model In such cases what you need to do is to create a new population using the newest model and then import your previous work into the new population using this import button on the properties menu However special care must be paid since import operation will import agent memory variable specifications for only those that exists in your latest model Therefore if there are newly added memory variables or agents it is up to you to complete the specifications for such new variables and agents A restriction of import operation is that number of regions in the current population must match that of the population you are importing from Therefore even if you plan to change number of regions yo
42. reNeighbour 10 getSelfVar region getAgentVar id If the selected variable is a struct or a list its elements can be accessed using a syntax similar to the examples given for getSelfVar e g getAgent Bank getAgentVar interestrates The agents are selected randomly by the current implementation of the above functions in PopGUI to prevent procedural bias in selection Exclusive selection If one wishes to prevent others from selecting the same agent it is possible to do so by setting the exclusive flag to 1 in getAgent variants i e getAgentRegional Employee exclusive 1 Once you do this the selected employee will never be selected again However one must be careful since it is possible depending on the composition of the population that such conditions may not be satisfied once all agents are taken Selecting multiple agents at once Two similar functions allow one to select all agents that match the criteria instead of only one get AllA gentsRegional agentname conditions get AllA gentsGlobal agentname conditions The variants accept same arguments as getAgentRegional and getAgentGlobal However what is returned is a list of agents These functions has an optional argument which can be used to turn of random shuffling of the returned agent list For example getAllAgentsRegional Firm randomize 0 will always return the same list in the same order If one needs to
43. s Probabilities must add up to 1 0 Also you can combine any of the above in arithmetic expressions e g 2 rand 0 5 or rand 2 6 choice 1 3 6 etc Discrete versions of random distributions are automatically chosen by PopGUI In other words even if you use random 1 0 2 0 for a variable whose type is integer it will be automatically treated as random 1 2 by PopGUI The expressions you enter are essentially Python expressions which must produce a single value with the exception of array initialization see Section 7 4 5 If you have some knowledge of Python you can enter any valid expression combining the special functions provided by PopGUI While entering specifications the Syntax help button on the bottom of this screen can provide you a quick reference help as an alternative to this user manual Once you are finished entering specifications you can use the Validate button on this screen to ensure that your expressions are correct This validation is mostly a syntactic one and does not guarantee that you will be able to generate a population instance at the end That depends on various things including setting up of relations and its success cannot be determined before you actually proceed to create an instance 7 4 3 Deterministic initialization deterministic min max lambda function will assign values by taking one value at a time from inclusive range and applying function to selected value
44. sfy implementation and agile development needs For this reason modules are written using different computer languages C Python and uses various software and libraries GTK Qt R project Despite their indepen dence the modules in EURACE GUI can be invoked easily from within an entry point program called Control GUI ConGUI These independent modules all correspond to stages of the process of designing running and evalu ation multi agent simulations of complex systems Following is a brief description of each module in the EURACE GUI ConGUI ConGUI is a front end which facilitates invocation of different modules of EURACE GUI Chapter 2 Introduction 4 easily as part of a design pipeline XMML Editor Used for designing different types of agents their memory variables and behavior in addition to global constants and messages used in a multi agent system GXparser Used as a front end for validating agent models produced with XMML Editor by parsing them through FLAME s xparser PopGUI Used for creating populations of agents whose state variables and relations can be initialized according to given criteria ExpGUI Used for creating sets of experiments using agent models and populations then for executing them on a FLAME simulation engine VisGUI Used for visualization and statistical analysis of simulation outcomes The reader is recommended to read Chapter 4 which provides an overview of the multi agent
45. tat average_wage png Eurostat average wage the plot on home bulentCode Plots and reports will be stored a previously picked working folde nome bulent Code visgui exper Agent Eurostat 0 8 4 0 4 Previous simulation location is a home bulent Code visgui Eurad Frequency Memory no_households_skill_5 gt Iterations 0 xml 7469k 100 xml 55k 1000 xml 55k 1020 xml 55k 1 1040 xml 55k 1060 xml 55k 1080 xml 55k 1100 xml 55k 1120 xml 55k 1140 xml 55k 1160 xml 55k 1180 xml 55k 120 xml 55k 1200 xml 55k 1220 xml 55k 12220 xml 56k 1240 xml 55k 1260 xml 55k 1280 xml 55k no_households_skill_5 1300 xml 55k 3 meds 2png 12 mamn 4 D 300 350 m bulentEX61s Doc lI visgui tex Docum f screenshots File Br EURACE Data Visua Ee Figure 9 4 Visualization of Distribution of a Variable 9 3 1 Platforms VisGUI is developed and tested both in Linux and Windows environments The application will also be adopted for MAC platforms 9 3 2 External Packages The application is being implemented using Python2 5 and Qt4 All distributional statistics and time series analysis and inference analysis are being performed by employing and integrating RPy2 RPy has provided an efficient and practical Python interface to the R Programming La
46. ted within the con text of agent being evaluated for selection because of the inherent way how Python lambda func tions are evaluated This was the reason that functions such as subequals subsubequals were created 4 Using external replicators to create hude populations If you start PopGUI with r option the created xml files will have all agent ID numbers preceeded with the string REPLACE ID Chapter 8 Running Simulations through ExpGUI The experiment manager module is highly coupled with policy analysis module Policy experiments require comparison of outcomes of alternative simulations where the compared cases correspond to co variation of several parameters among their corresponding ranges ExpGUI facilitates creation of a series of simulation tasks via a GUI thus reducing errors in creation of policy experiments File Help General Experiment Properties Population File in Use C FP6 EURACE Dissemination FETO3 Demo Eurace_Model_Demo its_larger_population 0 pop Number of initial XML files s Number of batch runs per XML fie 16 Default number of iterations Simulation Executable Root directory for output Select Folder Description of the Experiment Experiment Design Type Name Description Value Min Max Step Usc Agent Snapshot Period pei None None None F 1 Snapshot 1 None None None None None None 6 5 None None None None None None 7 Location of the O xmi File New P
47. to perform analyses and visualizations of simulations data which are generated elsewhere Third this decoupling feature allowed us to develop the application upon existing GP1f licensed tools and software In return after a maturation stage of the application the package will be made available publicly on public repositories as a GPL package This will enable a con tribution back to the open source community and contribution from the developers from the community to extend and maintain it in the future 9 2 Features A set of features are developed and made available to designers The features are briefly listed below and some snapshots of their use are given accordingly to ease presentation e Importing Simulation Results The user is made able to import either result of a single simulation or a set of it If this option is opted the application reads in a sample iteration and GNU Public License see urlhttp www gnu org 2Such as urlhttp sourceforge net 30 Chapter 9 Analysis of results using VisGUI 31 gt Applications Places System 4 ZS a N R Bulent Ozel wil Mon Oct 19 16 26 EURACE Data Visualisation and Analysis Workspace i Cte File Tools Settings Visualisation Analysis Window Help tr EVE GSR eS 4 VisGUI Confirmation the plot on home bulent Code An SQLite database already exists in the folder Plots and reports will be stor
48. u must do so only after importing NOTE As new features are added to PoPGUI it may not stay backwards compatible For ensuring stability data structures used by PopGUI are given a version number If you try to open an old population with newer versions of the PopGUI there is a chance that it will fail with a warning about the versions The recommended way to handle such situations is to create a population from scratch and import memory variables as described above 7 3 STEP 2 Specifying population composition Next step is to set number of each type of agent in each region When you click Edit Regions button in the toolbar you will be presented a table whose columns are regions and whose rows are agents see Figure 7 2 You can enter the number of agents in this grid and then press update and close button if you want to save these changes Your entries must be either positive integers or zero 7 4 STEP 3 Specifying agent memory variables This is the most elaborate step of specifying populations in PopGUI When you press Edit memory variables toolbar button you will be presented with a detailed display of agents and their variables with several buttons at the bottom part of screen as shown in Figure Agents their memory variables and sub components of these memory variables are presented in the form of a tree parts of which can Chapter 7 Population Design with PopGUI 17 File Help Model Summary 1 Properties
49. ui Eurace Mode Finish Time Mon Oct 19 16 51 25 2009 mm ec En ME E bulent i winsz Em mean_wage double Computer gq base needed_capital_stock double Ea 03 actual_cap_price double mean_specific_skills double planned_production_quantity double production_quantity double planned_output double output double unit_costs double total_supply double production_costs double revenue_per_day double technological_frontier double cum_revenue double B4 out of stock costs double 411 5 malls_sales_statistics array mall_id i E a malls_sales_statistics array sales arr e ise E mal sales tsteslaray sleslar 6 quality double CI TI BL et price_last_month double y demand _capital_stock double ViSGUI Raw plot data exporter planned production costs double adaption production volume due to m Bl bulentex61s Doc gt visgui tex Docum screenshots File Br Figure 9 6 Exporting Raw Data of Plots 36 Chapter 9 Analysis of results using VisGUI gt Applications Places System 4 AS N Bulent Ozel a W ail Mon Oct 19 17 00 5 a EURACE Data Visualisation and Analysis Workspace File Tools Settings Visualisation Analysis Window Help koa 000900 4 ETE EE Log 9 SQLite database file s has
50. ures of the VisGUI component will not be available without R However if your intention is merely using a computer as a multi agent design workstation it is perfectly possible to install and run EURACE GUI without these two software packages 3 1 Installing required software Although EURACE GUI can be installed and run without the following software its use and features are limited without them Therefore it is recommended that the following are installed in this order R R statistics package can be downloaded and installed from http www R project org If you are using a Debian or Ubuntu Linux it can also be installed using the package r base Chapter 3 Requirements and Installation 6 FLAME First get the sources from http ccpforge cse rl ac uk After unzipping the sources you need to compile and install two components from FLAME tree libmboard and xparser Enter to these directories in the given order run make then make install 3 2 Installing EURACE GUI from binary distribution To get a copy of EURACE GUI installer go to http eurace cs bilgi edu tr and download the most recent binary package for your platform supported platforms are Windows Linux and other Unix variants Execute this file to complete the installation 3 3 Installing EURACE GUI from sources If you are working on Microsoft Windows MinGW is required to compile and install the EURACE GUI from sourse distribution It c
51. vernment_GREQAM model xml enabled eurace_model xml 15KB xml File Input model file Statistical Office_UNIBI model xml enabled headerh 448 KB h File grading File so itary functiona e its_larger_population File Folder 20 04 2009 05 20 14 library_fi E gil s TT RT NO d its_smaller_ population File Folder 20 04 2009 05 20 26 Reading xagent named Firm low_primes h TLKB h File 25 10 2009 16 40 52 xparser Version 0 15 12 Reading xagent named Household Reading xagent named Mall main c 640 KB c File 25 10 2009 16 40 52 Reading xagent named IGFirm memory c 971 KB c File 25 10 2009 16 40 52 banal zn Mei denmeli ya ill messageboards c 273 KB c File 25 10 2009 16 Reading xagent named Bank AAA 900 bytes c File 20 04 2009 05 20 00 y_library Reading xagent named Government IN ee O my_library_headerh 234 bytes h File 20 04 2009 05 20 00 Reading message named bank account update partitioning c 9KB c File 25 10 2009 11 Reading message named central bank account update process order graph dot 131 KB dot File 25 10 2009 16 5 a D N N N rules c 256 KB c File 25 10 2009 11 AAA A gt stategraph dummy pdf 54 KB pdf File 20 04 2009 05 20 00 Figure 6 1 GUI for XParser 12 Chapter 7 Population Design with PopGUI PopGUI is written for creating agent populations for EURACE Project simulations using FLAME multi agent framework The program reads already designed agent models described in FLAME XMML for
52. w Chapter 7 Population Design with PopGUI 19 7 4 1 Distinguishing simple and composite variables Next to each variable s name in the memory variable editing screen you will see a variable type Following are the varieties of these types e A few variables are special Currently these are the ones named id and region_id Users cannot enter anything for these variables The id is a unique identifier of each agent in the population for referencing purposes and it will be assigned sequentially during population instantiation This variable must exist for FLAME simulations to work region_id when used will contain an integer which is the number of the region in which the agent is placed e Some variables are STATIC or CONSTANT These appear in cases where their value is fixed in the model or taken from a constant The names of constants are defined in agent model but their values are entered in PopGUI using Edit constants toolbar button e Simple variables has a type which is one of int or double e Composite variables are actually C structs The names and structure of these come from agent model which is used when creating a new population When a variable is of a composite type its sub fields can be expanded and will be entered in separate boxes e Finally a variable can be an array of a simple of composite type For such variables users will enter a specification for the length of arr
53. y variables be collapsed or expanded by clicking on the branches Alternatively all branches can be expanded or collapsed using the Expand all Collapse All buttons at the bottom of the screen This display also have a grid structure since one must provide specification for memory variables separately for each of the regions However in most cases the specifications for different regions of the same variable are same or very similar For this reason a Region copy button is provided in this screen Using this feature you can copy specifications for one region to other regions and then make incremental changes to speed up your work As described above specifications of memory variables are elaborate and has two basis values and relations Although both are addressed in the language used for memory variable initialization or initform as we refer to them in EURACE different features are required and thus we will present them in separate sections below The syntax help button on the memory variables screen displays up to date reference on the initform language for your convenience As a note to users the initform language is based on dynamic interpretation capability of Python and hence it has a Pythonic syntax However it is not necessary at all for users to have any experience in Python language although a little bit can help if one wants to use lambda expressions for extending the initform language as will be shown belo
Download Pdf Manuals
Related Search