Thursday, June 18, 2015

explination

So since Monday i have not been making post. This single post will encompass all the events up to this point...
---------------------------------------Monday------------------------------------- (^_^)
Ran simulations for point .6-.8 mutation rates on 20 threads. Also tested out continue class, all done on either imp or hornet... both are an 8 bit so the results could be comparable.
--------------------------------------Tuesday--------------------------------------(n_n)
The continue class has a bug, I knew because I put it there... it had a good property... all I need to do is find a way to save the property but not the bug.
The test results were coming in but all the results from hornet where 0.0 or -0.0. nearly all the results from imp where 1.0E-4. Dr. Remy and I walked through what could be happening and ran a another test with, to test the machine, so the same run was ran on joey, imp and hornet. And I was to manually run the test again on joey.
------------------------------------Wednesday-------------------------------------(-_-)
Results came in and hornet for what-ever reason only returns 0.0s, imp works for about 1 generation and the finesses return a default value. joey was the only one that performed okay. So I ran .6-.8 again on joey, .6 and .7 I used 10 threads but for .8 i used 20... just to see if it was a work load thing.
later that day I came back and all where running. .8 had a weird pattern where everything was okay but the second to last 10 finesses in the entire generation was a time out default, in each generation. I kept it going to see the effects,
----------------------------------Thursday-today----------------------------------(U_U)
for some reason, there is timeout values popping up here and there in every run. in the 20 threads the time out happens in the third run. in the 10 threads time out happens in the fourth run. This means that I will have to do some alterations to the code to allow for a longer wait time, or run  the evaluations with 10 threads which kinda defeats the purpose of a quick evaluation. Anyway Im just running the evals in small batches to see if the results are good

No comments:

Post a Comment