finalization

I think thats it.  I have had fun here at Clemson. I have to add some other little stuff to the presentation and will be working on another paper but I am going home.

plication

... its a word, look it up... So I wanted to write a letter to a friend, but I didn't have a letter, so I wrote the letter and thought i could try to make one. so I did, and It fits the letter really well. Look at the pictures

going from left to right.

line up letter to the center of paper

fold edges over letter

put letter on bottom of paper and fold to make crease. 

open paper and make triangle fold so that the lips above the line are close and the lips below open

fold paper closed with letter inside the closed lips.

close open lips on back side.

fold bottom of paper up over the lips on the backside, a flap should remain

fold flap of paper into pockets.

you have an origami envelop customized for your letter.

This works best if letter is thinner then actual envelop 

inculcation

So I have been using some cool programs like screen and ... um you know how when some one asked you what you do and then you instantaneously forget everything you do *_*... &. Im going to lay out a small description of them and some ways to use them. these are for terminals
Screen-----
allows for a user to have multiple screens in one terminal, and when terminal is closed no program progression is lost.  I all ready did a small description earlier... look for it.

&---
allows for a program to be done in the background.  to use just follow the command up with & and your done.

bg---
sends thing to background

fg -------
brings thing from the background

ctrl+z -------
stops program

git -----
git is like svn... kinda.  its a repository.... not to be confused with respiratory.... git has some follow up programs.
 ---add
 adds file to repository wait list
---help
  shows all the stuff git can do
 ---commit
  ensures stuff in wait list is going to go to repository
 ---clone
 copies repository to local machine
 ---push
  pushes all files on wait list to repository
--pull
 brings new files into local directory


there's others but I used these the most.

documentation

This is sort of a catch up post. On Monday, the paper was being finalized.. fixing all the pictures and sorting to find mistakes.... I had some trouble with compiling that paper with latex. I used a online editor called overleaf, and I was great until the compiler timed out, then I tried a offline compiler that I downloaded... didn't work... I tried another offline program plug in to eclipse... didn't work. Then I tried a different online editor.... nadda . So Dr Remy did some magic and now it compiles nicely.  On Tuesday I asked Peter and another guy who works in the lab...Brandon... to help me clean up my UI. It wasn't to bad. To run it the user needs some sort of background knowledge so I made sure the read-me was up and "nice". Today is going to be some backlog blog posts and two classes to help get around for the GA. they seem small so no biggy. Then at thing at 12 so Ill go. And after that, that's it. Tomorrow I will do a final presentation and be off to El Paso.

teleportation

well more like really fast transportation but... Last night i was able to run 21 runs from clemson to actual utah. It was quit interesting that I was able to run 21 runs back to back. Anyway I was able to chart the data and got some interesting results. these results wont make sense with out context though. The network runs from last week where also charted... those pics should be a bit lower or a post back... Those depicted the best individuals ran from joey to joey and joey to clemson Utah. The hypothosis was that a good individual will be good... in relation to the others of that population... regardless of the topology that the individual is ran on. these new charts which are the same individuals ran in actual Utah support that .... lookee.

 I know you can't see anything. All  the fitness where really bad. The highest fitness was like 730. out of 15000. But take into consideration how hard it is to balance an egg on a spoon, its not that hard, but now make that spoon a couple to hundred miles long and the story is different.  I redid  the chart with so that the highest y value would be 1000 and got a better picture.

these are the same just with different zooms... i guess is the word... But the picture is quite similar to that of the network run and the local run. 

accentuation

...light out... I spent the morning trying to figure out a way to single out the fitnesses of the most fit of the first generation using python. Doing so will answer some questions...like how good can one individual be? or what effect does the topology that the system is run on have on the same individuals.... Because the way my program is set up, so long as no files have been deleted all runs will start off with the same first generation... in one way or another, if the user chooses different seeded rate then the numbers will shift, but for all the experiment all the parameters where the same except for the mutation rate... Because of this all the runs have the exact same first generation. Except for the fitnesses all the logs look like the first generation has been copied and pasted. The program was kinda hard for me. In the end python was smaller, only like 43 lines of code and the first half was just imports. Arrays and indices are a little complex for me to use fluidly right now but im getting there. Anyway I made charts of 2 different typologies and the results are interesting.. to me at least .

Joey to Joey 

Joey to Joey

Joey to Utah

Joey to Utah

The "Joey to Joey" plots show the same trend. the first box... actually both one and two have the same look to them, so does three... There is little difference in either "joey to joey", this make sense because it is the same individuals on the same topology. But in terms of meaning the mean fitness is really close. There are bound to be stragglers but the fitness tends to be the same... or at least in the same area. The trend can be seen when comparing both "Joey to 'Utah'". But when it comes to comparing JTJ... Joey to joey... and JTU... joey to utah... there is a difference. In JTU all the boxes are notably smaller then there correspondents in JTJ, meaning that across the internet the fitness of an individual is more likely to be close to the true fitness.

telecommunication

I've been slacking, so this post will be half friday, the weekend and monday. 

Friday---------------------

I was able to make my program speak to me... 

Whenever the program has more then 10% defaults... like 0.0 or .1928 etc... my program will send me this txt... it can also send emails...

there's still some problems but those are being fixed, for example the user need to input the phone number in a form of an email, and the system cant send the email until it is finished running, but i have some ideas of how to go around that.

Weekend--------------------------------

During the weekend I finished the run with a mutation rate of .15. I dont think i mentioned this.

backtrack ----

there was a problem with the Utah run so the internet run was moved to a closer network that had the same topology of Utah. I ran a set of simulation but in .15 mutation. Then there was the problem of I didn't have a .15 mutation run from joey to joey or a .1 mutation run from joey to "Utah", so that is what was being done.

----

So that was the weekend. just getting runs done and stuff. One point that I am happy about is that with the net catching all the defaults in place I dont have to go over the log files to make sure the run didn't collapse.... less work .. Laziness Prevails.

Monday---------------------

on Monday I  had to finish one last little set of runs and bring the data together. There was an interesting trend. ...Looky

mutation rate 1 joey to joey

mutation rate .1 joey to "Utah"

mutation rate .15 joey to joey

mutation rate .15 joey to "Utah"

both of these charts show that when using the internet as a medium the simulation have a higher amount of good individuals.  One of the papers I read written by Zach McNellis and Dr. Remy suggested that not enough noise and the performance was poor, but enough noise and the performance was good. Like the Goldilocks zone... kinda anyway... Over the internet there is noise, point blank so the program works with that. from joey to joey, if someone runs something then the program freaks out and performance drops.

acquisition

Right now as of 2:50, I am both happy and tired. I was able to finish my net for catching defaults. To do so I needed to make my own exception, and a new class to hold all the defaults that the system has. I don't know all the defaults and as time went on I discovered new defaults and learned others where not. So the class that holds the defaults is dynamic. the one thing that I'm really happy about is that i was able to get the system to send a message when something is caught. If a person where to run using Presentor... is was changed from PresentorSlim around version 2... then if the system gets too many defaults then it stops, that's it. But if a person used GenAlgo... which is getting better... then GenAlgo will not only stop the system, the program will also print out a statement that the program stopped on the terminal but also send out an email or text message. Its not what i wanted exactly but the job gets done.  The one problem is that this makes GenAlgo platform dependent and can only send the message out through a Linux based system or Ubuntu... which is kinda the same but anyway... I'm happy with the results. I used a test directory so I could mess with code and not have to worry so now all i have to do is transfer over the files...

disconnection

Yesterday was a two part work. Some of the runs where corrupted so the data was skewed. These runs need to be redone... that is being worked on... I had this though before abut it didn't seem pertinent so it sat on the top self. But with collapse runs being passed through its important... at the very least convenient.... So I started constructing a net that sorts through the generations and if more then 10% of the population is a default then the system stops and a text is sent from the system to alert someone...The first part is done and installed, but the second part is still in the works.

The other part of yesterday was starting a run in "Utah"...which is a way of saying a network just like Utah in Clemson... For some reason whenever a manual fitness... the most basic form of fitness test... the value returned is zero, this is followed by the port being used to evaluate hitting an recursion error and crashing. I ran 5 simulations yesterday in the early afternoon and got no problems. I ran later in the day and all of the outcomes where defaults... either timeout or 0.0. So now I'm just continuing  with the net so this doesn't happen.

discombobulation

So I'm using latex for my paper and for some reason I cant get the references to work out for me. I didn't have a latex editor so I used overleaf. There's some problems with it but for the most part they are just a little unuser friendly.  So i downloaded texmaker and I dont know how to use it but im going to toy with it.

henicosaiconization

... okay, so it might seam like im grasping for words here but if  im correct that is a real word. Using the prefix hen-, which is one, and icosa-, which is 20. i get 21, these are Greek . using the icon root word i have picture. and -ization is the act of  making.  so all together i get the act of making 21 pictures/images, if im using the laws of English correctly...  Well I made a script that plotted all the finesses for each simulation. Here is the plots for mutation rate of one... idk why but when i run this script on joey i get this weird outcome.  Some have a weird kind of sine wave flow to them, and others are strait forward. the top one is the chart with all of the following in one. This top one show that even though the average growth levels off at around 25 generations, the system still explores. The other charts shows that each time is a different case but for the most part the system looks around.

manifestation

following this post will be others.
one will be some of the charts that where made
another will be some problems I had/having with latex/bibtex
and this one will be about today
so... uh oh, the following is an email

collaboration computation integration discombobulation frustration

As of now the genetic algorithm is ready for Peters program, the problem is that the system peters is using does not have any function for a feed back. The target zones are hard coded values, and the python back end cannot communicate those values back to the Ga. Peter is working of the issue, the change would allow for feed back, but its in the python. 

This problem is being worked on by the both of us. I am trying to make my program more user friendly, so thats on the agenda.

quadrilateralization

With all the data collected, I now need to make charts and stuff to understand what the data is saying. I have already made a python script that could make a box and whiskers plot given a file, so all I need/needed to do was to get all the log files fitness. I thought i could do this in java, but i know so little of python that I would give it a try. So far I have been able to get plots of each simulation. Now i need to get the plot that represents all the generation of each simulation... so like superimpose all of the simulations into one file. I also want to make it a one stop so I dont have to call anything else.  The challenge is that I have to add one array to the end of another array... I dont know how to do this... I think i can just call append, but I... well really i just have to toy with it...

pixelization

Just a promised, all the pictures.

area of work

screen shot of program running

screen shot of evaluation... and errors

shot of ball plate errors

elation

finally. all 21 samples of all 10 mutation rates is complete. now I just have to averege out the data and plot it. yah. and it only took .....emm well im done any how.

Ps I have screen shots of computers i was using but its on the macs in the office...and its raining so i will get them later...

perpetuation

right now it is about 12:00 and I still have not been able to get solid good simulation. Starting at 10 in the morning I was just running the computers. after every run I would reset the evals and ballPlates. The problem is that after x amount of time...x being some number based on the machines feelings... the runs would have crash...meaning that all the finesses would be either 0.0 or .1928... in some cases when I would start all 10 runs all would fail. This is what my day consisted of. running resetting, waiting, evaluating, sorting, deleting and repeating. In many cases only one run would come out with out more then 10 % collapse per generation. I would not it was based on the mutation rate...more of the machines that I was using. I found that the simulations on joey25 and up would last. my guess is that they are used less then the rest so they last longer. In any case I'm going to leave the runs now going and check in the morning. up to this point I have .1-.8 and I'm just waiting on .99 and .9.

Ps I used .99 because my GenAlgo will not take any mutation rate out of 0-1 so i though .99 would be really close to 1

revitalization

It was discovered that the link to the Utah computer was rejecting my/ Clemson`s request, so that was put on the back burner. Yesterday and today are the same thing. I am running the mutation runs again, but with 40 threads 40 individuals and 50 generations. If all is okay I can sneak in three simulations before the machine starts to pop zeros. The cause is still not  found, but there is a correlation to the work done by the machine. But in any case I'm just finishing up the simulation. In my GenAlgo there's some bugs... not stop the presses type but still... but there is also this nifty... yes, yes i did use that word... design that I can rerun a simulation with a simple yes. So all I have to do is restart the evals and the ballPlates... laziness prevails...