TB of RAM. I think I'm going to start a GoFundMe campaign to buy one for myself :)
2- The server's cost is about $13 an hour. I get free access to supercomputer through my university and xsede.org because I earned an NSF Honorable mention last March, however, the supercomputers available through both resources are a little complicated for me to use, as opposed to the one available from amazon that has Microsoft server 2012 already installed.
3- I wanted to run 400 annual glare simulations for 400 different views.
4- I tried a to perform annual glare simulation for one view on my Dell XPS that has Intel Core i7-6700HQ processor and 16GB of system memory. The simulation took 2 hours to complete. Radiance parameter ab was set to 6.
5- I wanted to obtain the batch file for each view so I can run them on the server. So I used the fly component to run all 400 simulations and closed the cmd windows, that wasn't bad ( for me at least) because I asked my son to this job for me, he was just glad to help me :)
6- I created one batch file using this cmd command:
dir /s /b *.bat > runall.bat
This created a file with the path to each .bat file. I edited this file in Notepad++ to include the word "start" at the beginning of each line. This was done using the "find and replace" dialogue box.
7- I split my newly created batch file into 3 batch files, each one has about 130 file names and " start" before the file names.
8- installed radiance on my server
9- Ran the first batch file on the server, this started 130 cmd windows performing my simulations, CPU usage was anywhere between 90% to 100% and about 105 GB of RAMs were used.
10. It took about 5 hours to complete all 130 simulations, I expected to run all in 2 hours but can't complain because this would've taken about 260 hours to run on my laptop. After the simulations done I ran the second and then the third batch files ( total of about 15 hours).
11. I got 400 valid dgb files. Couldn't be happier!
…
ss lots of questions,Hope guys show me some more different ways to figure out thoes kinds of problems,Thanks.
That is a construction project,the balconies should be overhang between 1 to 3 meters.
Program A is a patten consist of increasing balconies as the floors get upper.(In the picture is 29 at the first floor and ended with 2 more balconies for each floor, )Each part for a different floor,the twelfth floor have 29+(12-1)*2=51 balconies.
Questions From A,
A1:How to use the {(series)} to creat this atrium,As the floors increase the number of the balconies change by arithmetic progression.
A2:How to control the angle of the balconies,both the angle with floor and the balconies ending part.
Program B is use line to shape the commercial atrium,program A is more small pieces of rectangles.The {(TweenCrv)} command.
Questions From B,
B1:How to draw random points between the 1 to 3 meters region of the balcony,And those point form a shape also belongs to that region.
B2:Use a curve or other ways to control the changing speed of each floors' balcony.Right now the balcony is a Linear change.
Thanks for your Help.
Q1:Is there a way in Grasshopper to control the model to Modulus,less different unit parts to build such a Atrium.(For Exanple,only use 900mm and 600mm two different width of the Glass railings to bulid the model A OR B)…
se (like in nature). the length of the sticks shall be controlled by the brightnessvalues of a picture. so the bend have to be controlled, too.
now we have several problems:
1. how can i map a hexgrid on a curved surface?
2. how can i adapt the grid to the dimensions of the surface (no overlap, no gaps to the bound)?
3. important
: to create the curved sticks, we use points on a line and we move some of them and then we want to connect the right points via interpolated curve to create each curved stick. now the problem is that the points have to been filtered in the right way. we know that we have to filter each list of points to the index values of the points. the number of index values is the number of hexgrid rows, so there are a lot and we can't use a list item for each one. it could be hundreds.
is there any opportunity to sort a list after the index values (first every index=0, then index=1, ...n)?
or is there any component which does a group of operations for n-times (n is the flexible number of index values) ?
4. how can i control the length and bend of the sticks via the brightnessvalues of a picture?
please help us. thanks.
german version:
In einem hexagonalen Raster soll sich senkrecht zu Oberfläche ein Stab im Mittelpunkt jedes Sechsecks befinden. Dieser soll sich ab einem gewissen (festgelegten) Punkt Richtung Boden biegen. Zusätzlich wird die Länge des Stabes zum Beispiel durch die Information eines Bildes gesteuert, so dass auch die Biegung, je nach Länge, geregelt werden muss.
Wir haben ein Hexagonales Grid (HexGrid) erzeugt und in jeden Mittelpunkt eine Linie senkrecht zum Grid erzeugt, aus der wir uns Punkte mit CurvePoint ausgeben lassen. Der letzte ist verschoben, um eine Biegung zu simulieren. Um die Punkte zu einer interpolierten Kurve zu verbinden, müssen sie nach dem Index sortiert werden. Gibt es eine andere Möglichkeit, als jeden einzelnen Indexwert über ein ListItem herauszufiltern (Da die Rasterung flexibel einstellbar sein soll, entstehen n Indexwerte)? Oder kann man eine Liste nach den Indexwerten, also nicht nach den Punkten, sortieren?
Und wie kann man über Bildhelligkeitswerte die Länge der Stäbe und damit auch die Biegung steuern (ein kurzer Stab biegt sich weniger als ein langer Stab)?
Gibt es die Möglichkeit ein hexagonales Raster auf eine gekrümmte Fläche zu mappen?
Und wie passt man ein solches Raster (HexGrid) in eine Fläche mit definierten Maßen ein, ohne dass das Raster an den Rändern übersteht oder die Fläche nicht vollkommen ausfüllt?
danke.…
Added by doro hamann at 7:34am on December 20, 2011
what they really mean by that, as in what buttons to push, so I assume it's a Windows Path entry?
2.) Modify PATH
Add the install location on the path, this is usually: C:\Program File\IronPython 2.7
But on 64-bit Windows systems it is: C:\Program File (x86)\IronPython 2.7
As a check, open a Windows command prompt and go to a directory (which is not the above) and type:
> ipy -V PythonContext 2.7.0.40 on .NET 4.0.30319.225
Tutorial on setting a Windows environmental variable (path):
http://www.computerhope.com/issues/ch000549.htm
But this fails to point out that path contains many entries already separated by semicolons so if I merely add a new variable called "path" it's likely that I will destroy existing program function. There's no info on how to just tack on another entry, and the Windows 7 edit box doesn't even show the whole collection, but one item (!), so I copied the existing path into a text editor to see the whole collection successfully and added the C:\Program Files (x86)\IronPython 2.7 entry after an added semicolon, correcting for an Enthought page typo of no 's' on the end of "Program Files". I also checked the others and many pointed to old missing directories so I deleted those entries.
...and the test fails and "ipy" is not recognized as a command, even though the path now shows up using "path" in the Windows CMD window, that is if I copy all by right clicking and pasting the stuff into a text editor to really view it all. I can run it from the source directory just fine.
The rabbit hole was indeed deep. Using the Task Manager (control-alt-delete) to kill Explorer and then Run in the menu to restart "Explorer," along with restarting the Windows CMD window however, worked. I can now invoke Iron Python ("ipy") via command line from any directory. For the "path" I edited path in the System Variables and not the User Variables. No, you don't have to type that whole crazy line above just to test the path variable, just "ipy" (and control-Z to quite IronPython) in the CMD window invoked by typing "cmd" into the Start menu search box.
From the CMD line this step did work fine:
3.) ironpkg
Bootstrap ironpkg, which is a package install manager for binary (egg based) Python packages. Download ironpkg-1.0.0.py and type:
> ipy ironpkg-1.0.0.py --install
Now the ironpkg command should be available:
> ironpkg -h(some useful help text is displayed here)
But of course Step 4 fails, giving pages of what seem to be error messages;
C:\Users\Nik>ironpkg scipy
Traceback (most recent call last):
File "C:\Program Files (x86)\IronPython 2.7\lib\site-packages\enstaller\utils.
py", line 92, in write_data_from_url
File "C:\Program Files (x86)\IronPython 2.7\Lib\urllib2.py", line 126, in urlo
pen
File "C:\Program Files (x86)\IronPython 2.7\Lib\urllib2.py", line 397, in open
File "C:\Program Files (x86)\IronPython 2.7\Lib\urllib2.py", line 509, in http
_response
...
Why can't I just download Numpy as a normal file and thus also have it easy for other users to install it when they use my scripts? This is just crazy and lazy. The Enthought developer has turned this into a computer game, with a missing registration link and then the last step spits out errors with utterly no information on how to fix it manually.
This Step 4 error is covered here:
http://discourse.mcneel.com/t/trying-to-import-numpy-in-rhino-python-but-im-getting-this-error-cannot-import-multiarray-from-numpy-core/12912/16…
Added by Nik Willmore at 2:36pm on October 11, 2015
he picture (4).
Previously, I had a problem with generating intersections between the two directions of the beams, but a colleague helped me by extending beams, so there was no problem with lines of intersection. But this solution has generated curl (5) at the highest vertex geometry, which I ignored in order to repair it before printing, perhaps this mean my problem with my beam spread properly. Only when the beams is 19, does not jump no problem, but I still can not distribute them properly.
(1)
(2)
(3)
(4)
(5)
I tried to show as simply as possible by removing or signing my code in GHX file.
Thank you in advance for your help
…
.
Things have been working swimmingly in many areas of the plugin, but one particular problem has been tough to solve. I have two components that are trying to read/write to the same memory at the same time, causing Rhino exceptions and crashes.
The conflicts appear to be happening between two components -- one is a "Layer Events Listener" that reports essentially what type of layer event just happened. The other is a "Set Layer Visibility" component that toggles the visibility of a list of layers.
The code:
public class LayerTools_LayerEventsListener : GH_Component { /// <summary> /// Initializes a new instance of the LayerTools_LayerListener class. /// </summary> public LayerTools_LayerEventsListener() : base("Layer Events Listener", "Layer Listener", "Get granular information about the layer events happening in the Rhino document.", "Squirrel", "Layer Tools") { }
/// <summary> /// Registers all the input parameters for this component. /// </summary> protected override void RegisterInputParams(GH_Component.GH_InputParamManager pManager) { pManager.AddBooleanParameter("Active", "A", "Set to true to listen to layer events in the Rhino document.", GH_ParamAccess.item, false); pManager.AddTextParameter("Exclusions", "E", "Provide a list of exclusions to stop reading specific events (Added, Deleted, Moved, Renamed, Locked, Visibility, Color, Active).", GH_ParamAccess.list); pManager[1].Optional = true; }
/// <summary> /// Registers all the output parameters for this component. /// </summary> protected override void RegisterOutputParams(GH_Component.GH_OutputParamManager pManager) { pManager.AddBooleanParameter("Initialized", "I", "Whether the listener changed from passive to active.", GH_ParamAccess.item); pManager.AddTextParameter("Document Name", "doc", "Name of the Rhino document that is changing.", GH_ParamAccess.item); pManager.AddTextParameter("Layer Path", "path", "Path of the modifed layer.", GH_ParamAccess.item); pManager.AddIntegerParameter("Layer Index", "ID", "Index of the modified layer.", GH_ParamAccess.item); pManager.AddIntegerParameter("Sort Index", "SID", "Sort index of the modified layer.", GH_ParamAccess.item); pManager.AddTextParameter("Event Type", "T", "Type of the modification.", GH_ParamAccess.item); pManager.AddBooleanParameter("Added", "A", "If the layer has been added.", GH_ParamAccess.item); pManager.AddBooleanParameter("Deleted", "D", "If the layer has been deleted.", GH_ParamAccess.item); pManager.AddBooleanParameter("Moved", "M", "If the layer has been moved.", GH_ParamAccess.item); pManager.AddBooleanParameter("Renamed", "R", "If the layer has been renamed.", GH_ParamAccess.item); pManager.AddBooleanParameter("Locked", "L", "If the layer locked setting has changed.", GH_ParamAccess.item); pManager.AddBooleanParameter("Visibility", "V", "If the layer's visibility has changed.", GH_ParamAccess.item); pManager.AddBooleanParameter("Color", "C", "If the layer's color has changed.", GH_ParamAccess.item); pManager.AddBooleanParameter("Active", "Act", "If the active layer has changed.", GH_ParamAccess.item); }
/// <summary> /// This is the method that actually does the work. /// </summary> /// <param name="DA">The DA object is used to retrieve from inputs and store in outputs.</param> protected override void SolveInstance(IGH_DataAccess DA) { bool active = false; List<string> exclusions = new List<string>();
DA.GetData(0, ref active); DA.GetDataList(1, exclusions);
RhinoDoc thisDoc = null;
bool initialize = false;
string dName = null; string activePath = null; int layerIndex = -1; int sortIndex = -1; string eventType = null; bool added = false; bool deleted = false; bool moved = false; bool renamed = false; bool locked = false; bool visibility = false; bool color = false; bool current = false;
if (active) { thisDoc = RhinoDoc.ActiveDoc;
initialize = (!previouslyActive) ? true : false;
RhinoDoc.LayerTableEvent -= RhinoDoc_LayerTableEvent; RhinoDoc.LayerTableEvent += RhinoDoc_LayerTableEvent; previouslyActive = true;
} else {
RhinoDoc.LayerTableEvent -= RhinoDoc_LayerTableEvent; previouslyActive = false; }
if (ev != null) { dName = ev.Document.Name; layerIndex = ev.LayerIndex; eventType = ev.EventType.ToString();
if (!exclusions.Contains("Active")) { if (ev.EventType.ToString() == "Current") { // active layer has just been changed current = true; }
}
if (!exclusions.Contains("Moved")) { if (ev.EventType.ToString() == "Sorted") { // active layer has just been changed moved = true; }
}
if (!exclusions.Contains("Added")) { if (ev.EventType.ToString() == "Added") { // layer has just been added activePath = ev.NewState.FullPath; added = true; }
}
if (!exclusions.Contains("Active")) { if (ev.EventType.ToString() == "Deleted") { // layer has just been added
deleted = true; } }
if (ev.EventType.ToString() == "Modified") { // layer has been modified activePath = ev.NewState.FullPath;
//skip sortindex eventType = ev.EventType.ToString();
if (ev.OldState != null && ev.NewState != null) { if (!exclusions.Contains("Locked")) { if (ev.OldState.IsLocked != ev.NewState.IsLocked) locked = true;
} if (!exclusions.Contains("Visibility")) { if (ev.OldState.IsVisible != ev.NewState.IsVisible) visibility = true; }
if (!exclusions.Contains("Moved")) { if (ev.OldState.ParentLayerId != ev.NewState.ParentLayerId) moved = true; }
//if (ev.OldState.SortIndex != ev.NewState.SortIndex) moved = true; if (!exclusions.Contains("Renamed")) { if (ev.OldState.Name != ev.NewState.Name) renamed = true; }
if (!exclusions.Contains("Color")) { if (ev.OldState.Color != ev.NewState.Color) color = true; } }
} }
DA.SetData(0, initialize); DA.SetData(1, dName); DA.SetData(2, activePath); DA.SetData(3, layerIndex); DA.SetData(4, sortIndex); DA.SetData(5, eventType); DA.SetData(6, added); DA.SetData(7, deleted); DA.SetData(8, moved); DA.SetData(9, renamed); DA.SetData(10, locked); DA.SetData(11, visibility); DA.SetData(12, color); DA.SetData(13, current);
}
static bool previouslyActive = false; Rhino.DocObjects.Tables.LayerTableEventArgs ev = null;
void RhinoDoc_LayerTableEvent(object sender, Rhino.DocObjects.Tables.LayerTableEventArgs e) { ev = e;this.ExpireSolution(true); }
And for the layer visibility component:
public LayerTools_SetActiveLayer() : base("Set Active Layer", "SetActiveLayer", "Set the active layer in the Rhino document.", "Squirrel", "Layer Tools") { }
/// <summary> /// Registers all the input parameters for this component. /// </summary> protected override void RegisterInputParams(GH_Component.GH_InputParamManager pManager) { pManager.AddBooleanParameter("Active", "A", "Set to true to change the active layer in Rhino.", GH_ParamAccess.item, false); pManager.AddTextParameter("Path", "P", "Full path of the layer to be activated.", GH_ParamAccess.item); }
/// <summary> /// Registers all the output parameters for this component. /// </summary> protected override void RegisterOutputParams(GH_Component.GH_OutputParamManager pManager) { pManager.AddIntegerParameter("Layer ID", "ID", "Index of layer that has been activated.", GH_ParamAccess.item); pManager.AddBooleanParameter("Status", "St", "True when the layer has been activated.", GH_ParamAccess.item); }
/// <summary> /// This is the method that actually does the work. /// </summary> /// <param name="DA">The DA object is used to retrieve from inputs and store in outputs.</param> protected override void SolveInstance(IGH_DataAccess DA) { bool active = false; string path = "";
if (!DA.GetData(0, ref active)) return; if (!DA.GetData(1, ref path)) return;
int layer_index = -1; bool status = false;
if (path != null) {
Rhino.RhinoDoc doc = Rhino.RhinoDoc.ActiveDoc; Rhino.DocObjects.Tables.LayerTable layertable = doc.Layers;
layer_index = layertable.FindByFullPath(path, true);
if (layer_index > 0) { // if exists RhinoDoc.ActiveDoc.Layers.SetCurrentLayerIndex(layer_index, true); status = true; } }
DA.SetData(0, layer_index); DA.SetData(1, status); }
Now originally I was getting exceptions when changing multiple layers' visibility properties, which would cause the Event Listener to fire and try to read the Visibility property before the memory has been released by the Set Layer Visibility component. That led me to add an "Exceptions" input, that would allow me to disable the reading of Visibility events at the source in the Layer Events listener. That helped me manage about 95% of the crashes I was getting, but I still get strange crashes in other event properties, even when that property shouldn't be affected. For instance, I am getting a crash here on the Name property in the event from the delegate function, even though I am only changing Visibility at any one time:
I have a few ideas but they all seem pretty hacky. One is to try to set a flag that is readable by any component in the plugin -- so that the event listener can see if a "set" component is currently running and abort before causing an exception. The other is creating a delay in the event listener, somthing like 200ms, to allow any set components to finish what they are doing before reading the event. Neither seems super ideal.
Any ideas?
Thanks,
Marc
…
currently within a fake euphoria framework - blame China/UAE) and a potential decision about doing/developing this or doing that … well …anyway … read and enjoy.
AEC matters: The good, the bad and the ugly.
The bad news: Rhino is NOT suitable for the job (although some use it … but only in the sense that people use Modo for the so called “hard modeling”). By job I mean things up to shop drawing level + specs + you and me (we call it Final level) – nothing to do with sketches and outlines of some abstract “schematic” topology.
The ugly news: The so called Design-Construct approach gains exponentially momentum especially in countries the likes of China/UAE/BRICS (95% of the whole AEC activity worldwide happens there). DeCo means: AEC engineers deliver some kind of study in a preliminary level and the main contractor splits (outsourcers) the job and assigns the study completion AND the construction to various sup-constructors. That thing appeared first – in a large scale - in Dubai 15 years ago. This means that the era of Sergio Pininfarina is over and out: welcome anonymous Toyota designer. In plain English: days of construction corporations fast replacing practices. Dead men walking.
The good news: All AEC related apps (Revit, AECOsim, Allplan and the likes) are in a lethargic state as regards the brave new world (based on the archaic level driven organization schemas etc etc). Of course they all claim the exact opposite and point that support BIM (nobody mention PLM) better than the other guy. But the 21st century – helped by 2 forthcoming unavoidable crisis (a) about shortage of water (b) about transition from carbon to hydrogen economy – isn’t about bureaucracy: think cost/resources optimization and “fitness” rather than China/UAE type of liquid trend. Days of euphoria fast approaching the Wall.
Top to bottom and visa versa.
Old days Titans (Oscar, Mies, Walter, Pierre Luigi, Frank, Eero, blah blah) outlined things (mostly using crayons) and the rest were struggling to translate these in reality in an one way vector like process : Top to bottom that is. These days the inverse gains momentum : when in the whole consider the part … validate … redo … validate … redo. This means bottom to top geared with top-to-bottom. In plain English : child imposes rules to parent and parent imposes rules to child. This means classic MCAD feature/history modeling (CATIA/NX/MS). This is something that Rhino can’t do (not to mention that Rhino is a surface modeler – a rather critical fact).
The parts that are bigger than the whole.
Go there ( http://www.behance.net/gallery/2885057/a-myriad-of-cables) are inspect the whole thing: it’s a parametric nightmare made with the other guy (Generative Components – slower than a Skoda + bugs + why bother?). But the whole (masts and membranes and the likes) means nothing here: focus to the details that are critical for connecting this with that. Complex feature driven solids that are made with internal (on a per se basis) parameters (like fillets required for casting or radius for cable anchoring) whilst they comply with external rules/parameters (cable angles, topology clash issues etc etc). So the whole outlines possibilities … and the part either can follow…or the part must change…or the whole must change. Can you do that with GH/Rhino? And if not what’s missing? (lot’s of things to be honest).
Some other "similar" things:
The narrow picture.
I agree with what others already said and with pretty much all Ola’s points – especially the visual drag-and-drop path mapper (i.e. a visual data manipulator so to speak) and the enable/disable components in groups capability.
Some other suggestions:
A multi canvas capability. As things are right now…it’s like working in Rhino in one view (rather unsuitable I guess). In fact …since overlapping views they don’t work in Rhino…well…you know, he he.
A working auto profile arrangement capability (non twisted Loft/Sweep and the likes). Worth 1Bn dollars that one.
Ability to locate components that caused this or that in the Rhino view: meaning a 2 way communication approach : GH makes things happen in Rhino and things can indicate their cause in the GH canvas.
A robust collection of components that bake stuff in nested blocks (emulating some primitive assembly/component way of thinking). Why may you ask? Well … the whole objective is to talk to CATIA (via STEP) don’t you agree? CATIA makes things happen in real-life not Rhino.
A robust collection of components that can create real-life parametric tensile membrane solutions (get some inspiration from FormFinder: useless because it’s academic but good to point the way). Membranes (and geodesics) are the future.
I could continue at infinitum but IMHO the big picture is worth 12345,67 “focused” GH improvements.
May the Force (the dark option) be with us all.
…
lly it should not make much of a difference - random number generation is not affected, mutation also is not. crossover is a bit more tricky, I use Simulated Binary Crossover (SBX-20) which was introduced already in 1194:
Deb K., Agrawal R. B.: Simulated Binary Crossover for Continuous Search Space, inIITK/ME/SMD-94027, Convenor, Technical Reports, Indian Institue of Technology, Kanpur, India,November 1994
Abst ract. The success of binary-coded gene t ic algorithms (GA s) inproblems having discrete sear ch sp ace largely depends on the codingused to represent the prob lem variables and on the crossover ope ratorthat propagates buildin g blocks from pare nt strings to childrenst rings . In solving optimization problems having continuous searchspace, binary-co ded GAs discr et ize the search space by using a codingof the problem var iables in binary st rings. However , t he coding of realvaluedvari ables in finit e-length st rings causes a number of difficulties:inability to achieve arbit rary pr ecision in the obtained solution , fixedmapping of problem var iab les, inh eren t Hamming cliff problem associatedwit h binary coding, and processing of Holland 's schemata incont inuous search space. Although a number of real-coded GAs aredevelop ed to solve optimization problems having a cont inuous searchspace, the search powers of these crossover operators are not adequate .In t his paper , t he search power of a crossover operator is defined int erms of the probability of creating an arbitrary child solut ion froma given pair of parent solutions . Motivated by t he success of binarycodedGAs in discret e search space problems , we develop a real-codedcrossover (which we call the simulated binar y crossover , or SBX) operatorwhose search power is similar to that of the single-point crossoverused in binary-coded GAs . Simulation results on a number of realvaluedt est problems of varying difficulty and dimensionality suggestt hat the real-cod ed GAs with t he SBX operator ar e ab le to perform asgood or bet t er than binary-cod ed GAs wit h t he single-po int crossover.SBX is found to be particularly useful in problems having mult ip le optimalsolutions with a narrow global basin an d in prob lems where thelower and upper bo unds of the global optimum are not known a priori.Further , a simulation on a two-var iable blocked function showsthat the real-coded GA with SBX work s as suggested by Goldberg
and in most cases t he performance of real-coded GA with SBX is similarto that of binary GAs with a single-point crossover. Based onth ese encouraging results, this paper suggests a number of extensionsto the present study.
7. ConclusionsIn this paper, a real-coded crossover operator has been develop ed bas ed ont he search characte rist ics of a single-point crossover used in binary -codedGAs. In ord er to define the search power of a crossover operator, a spreadfactor has been introduced as the ratio of the absolute differences of thechildren points to that of the parent points. Thereaft er , the probabilityof creat ing a child point for two given parent points has been derived forthe single-point crossover. Motivat ed by the success of binary-coded GAsin problems wit h discrete sear ch space, a simul ated bin ary crossover (SBX)operator has been develop ed to solve problems having cont inuous searchspace. The SBX operator has search power similar to that of the single-po intcrossover.On a number of t est fun ctions, including De Jong's five te st fun ct ions, ithas been found that real-coded GAs with the SBX operator can overcome anumb er of difficult ies inherent with binary-coded GAs in solving cont inuoussearch space problems-Hamming cliff problem, arbitrary pr ecision problem,and fixed mapped coding problem. In the comparison of real-coded GAs wit ha SBX operator and binary-coded GAs with a single-point crossover ope rat or ,it has been observed that the performance of the former is better than thelatt er on continuous functions and the performance of the former is similarto the lat ter in solving discret e and difficult functions. In comparison withanother real-coded crossover operator (i.e. , BLX-0 .5) suggested elsewhere ,SBX performs better in difficult test functions. It has also been observedthat SBX is particularly useful in problems where the bounds of the optimum
point is not known a priori and wher e there are multi ple optima, of whichone is global.Real-coded GAs wit h t he SBX op erator have also been tried in solvinga two-variab le blocked function (the concept of blocked fun ctions was introducedin [10]). Blocked fun ct ions are difficult for real-coded GAs , becauselocal optimal points block t he progress of search to continue towards t heglobal optimal point . The simulat ion results on t he two-var iable blockedfunction have shown that in most occasions , the sea rch proceeds the way aspr edicted in [10]. Most importantly, it has been observed that the real-codedGAs wit h SBX work similar to that of t he binary-coded GAs wit h single-pointcrossover in overcoming t he barrier of the local peaks and converging to t heglobal bas in. However , it is premature to conclude whether real-coded GAswit h SBX op erator can overcome t he local barriers in higher-dimensionalblocked fun ct ions.These results are encour aging and suggest avenues for further research.Because the SBX ope rat or uses a probability distribut ion for choosing a childpo int , the real-coded GAs wit h SBX are one st ep ahead of the binary-codedGAs in te rms of ach ieving a convergence proof for GAs. With a direct probabilist ic relationship between children and parent points used in t his paper,cues from t he clas sical stochast ic optimization methods can be borrowed toachieve a convergence proof of GAs , or a much closer tie between the classicaloptimization methods and GAs is on t he horizon.
In short, according to the authors my SBX operator using real gene values is as good as older ones specially designed for discrete searches, and better in continuous searches. SBX as far as i know meanwhile is a standard general crossover operator.
But:
- there might be better ones out there i just havent seen yet. please tell me.
- besides tournament selection and mutation, crossover is just one part of the breeding pipeline. also there is the elite management for MOEA which is AT LEAST as important as the breeding itself.
- depending on the problem, there are almost always better specific ways of how to code the mutation and the crossover operators. but octopus is meant to keep it general for the moment - maybe there's a way for an interface to code those things yourself..!?
2) elite size = SPEA-2 archive size, yes. the rate depends on your convergence behaviour i would say. i usually start off with at least half the size of the population, but mostly the same size (as it is hard-coded in the new version, i just realize) is big enough.
4) the non-dominated front is always put into the archive first. if the archive size is exceeded, the least important individual (the significant strategy in SPEA-2) are truncated one by one until the size is reached. if it is smaller, the fittest dominated individuals are put into the elite. the latter happens in the beginning of the run, when the front wasn't discovered well yet.
3) yes it is. this is a custom implementation i figured out myself. however i'm close to have the HypE algorithm working in the new version, which natively has got the possibility to articulate perference relations on sets of solutions.
…