10:53 < mentekid> rcurtin: I've run some more tests for Corel, Phy, Miniboone and Pokerhand

10:53 < mentekid> I've packed all time data here, if you run it with matlab it plots everything so it gives us a better idea

10:53 < mentekid> http://pastebin.com/RtXWuKpn

10:55 < mentekid> it looks like it's all over the place - in some case there's improvement, in others it's worse

10:56 < mentekid> Do you think it might be better to keep this one stashed until I've implemented the tuning algorithm? So we can then check with realistic parameter sets and get more decisive results

10:57 < mentekid> I shouldn't say realistic... optimal might be a better word

11:40 < rcurtin> mentekid: you're right, that is not at all like the pattern I expected

11:41 < rcurtin> I think that your conclusion might be right... that we should wait until later to see if we can figure out how to apply this

11:41 < rcurtin> I'm going to spend a little while trying to think about why we are seeing such unexpected patterns here

16:04 < mentekid> rcurtin: sorry I had trouble with my computer. Yeah I don't understand this either, could it just be my machine?

16:05 < mentekid> I ran it at night so I wouldn't be using my computer so the time would be more reliable though, so I don't expect there to be a difference if you run it on a server

17:27 < rcurtin> yeah, let me try running the same thing

17:27 < rcurtin> did you have a nice script to do it?

17:30 < mentekid> no nothing particular, just added -v and piped everything to grep

17:31 < mentekid> "runLSH.sh": http://pastebin.com/FxuRfQhK

17:33 < rcurtin> ok, let me give it a shot

17:34 < rcurtin> actually it might be a little while, I am in a meeting now and can't set this up right

17:34 < rcurtin> so maybe it might be tonight before I can

17:34 < mentekid> the "query" files are only the top 10000 lines of each file

17:34 < mentekid> sure no hurry... I'll take a look on the vectorization thing we discussed