Page 5 of 9

Re: Proficiency learning optimizer

PostPosted: Sat Jan 31, 2015 12:19 pm
by Luaan
majtejn wrote:EDIT: The only idea I'd like to see is to be able to set a difficulty (1-10) for each item. So i can tell the program how hard is something to get.


Yeah, I've had that at one point - you could add a weight to any inspirational (using the inpirationals.dat file). It doesn't really work all too well - either the weight is too low to make a difference, or its so big you're not going to get that inspirational at all. Using the "Prune leaderboard" option is a lot more effective - you can see the runner ups and decide whether those inspirationals are more appropriate for you.

It would be handy to have something that also plans for longer learning, e.g. "Raise my threads & needles from 10000 to 20000 with this amount of inspirationals", but the magnitude of that problem is considerably larger.

It's also very tricky to combine optimizing for inspiration while optimizing for scarcity at the same time. I considered using a price-in-silvers as a scarcity metric, but it breaks down because some inspirationals are virtually free, while others are very scarce. And combining inspiration with cost in a reasonable way is very tricky. You can try tweaking the engine yourself, the sources are all there :)

Re: Proficiency learning optimizer

PostPosted: Sat Jan 31, 2015 12:46 pm
by Yes
Luaan wrote:It's also very tricky to combine optimizing for inspiration while optimizing for scarcity at the same time. I considered using a price-in-silvers as a scarcity metric, but it breaks down because some inspirationals are virtually free, while others are very scarce. And combining inspiration with cost in a reasonable way is very tricky. You can try tweaking the engine yourself, the sources are all there :)

That's what checking/unchecking is for. Silver price per study session is useless atm because there is no normal market price.

Re: Proficiency learning optimizer

PostPosted: Sat Jan 31, 2015 1:21 pm
by majtejn
The checking/unchecking works fine, but what if I have a Carpenter's square but I don't want to use it more than once in a session.
Maybe the program doesn't have to optimize for the difficulty, but there could be an option to sort possible solutions based on the total difficulty of the used items.
I imagine this by adding another column to both lists one for setting the diff, the other for displaying the summarized diff. This solution does nothing with the engine, that's the next step...

Re: Proficiency learning optimizer

PostPosted: Sat Jan 31, 2015 2:08 pm
by Kandarim
If you want to use your carpenter's square exactly once per session, just subtract whatever it gives from your goal per session.

Re: Proficiency learning optimizer

PostPosted: Mon Feb 02, 2015 12:58 am
by EnderWiggin
Luaan wrote:
EnderWiggin wrote:Can you add checkbox to multiply proficiencies by 100? That way it would be a bit easier to setup for raising proficiency levels.
I tend to use the mousewheel for that. The tricky part is that different mice (and mouse configurations) scroll by different amounts. Maybe I could hack the scrolling somehow to go 100 at a time always? So you'd just need to rotate your wheel two places to get to the next "level"?
What I meant is you write 11 in proficiency value and program treats it as 1100. That way it would be easier to make calculation for proficiencies raising - less 0's to type. You can even add "levelup" button to add 1 (or 100, depending on mode) to all non-zero proficiencies.

Re: Proficiency learning optimizer

PostPosted: Mon Feb 02, 2015 8:47 am
by majtejn
Added some little changes:
https://github.com/majtejn/SalemOptimizer

ZIP:https://www.dropbox.com/s/0lwrffybak3alzm/SalemOptimizer.1.6.2.zip?dl=0

-inspirationals now have difficulty
-solutions now have total difficulty (sortable)
-prof selection now scrolls by 100
-solution count now editable
-added Use All checkbox that will ignore selected inspirationals and act like all were selected
-calculation now goes on 8 threads (still hardcoded)
-added skill database for fast selection (doesn't work, something isn't right with the imported data)
Wiki guys out there, please make a table of skills like the one with inspirationals. I used a dumb crawler from import.io but that didn't work well.

Plans:
- to make skill selection right
- to make a new window with current proficiencies so in the window that comes up when you click on a solution could be a list of other proficiencies raised in the session
- maybe some way to optimize for difficulty too

Re: Proficiency learning optimizer

PostPosted: Fri Feb 06, 2015 5:25 pm
by trungdle
newest version sometime calculate wrong for me. can't replicate the problem as I spent all my insp on your tool and get all the skills I can in 2 days.
I both love and hate you now :|

Re: Proficiency learning optimizer

PostPosted: Sat Feb 07, 2015 3:30 pm
by RedVIII
trungdle wrote:newest version sometime calculate wrong for me. can't replicate the problem as I spent all my insp on your tool and get all the skills I can in 2 days.
I both love and hate you now :|

Maybe you used them in the same order which the optimizer, which don't order them to avoid the non-variety nerf.

Re: Proficiency learning optimizer

PostPosted: Sat Feb 07, 2015 3:33 pm
by alloin
Image

Re: Proficiency learning optimizer

PostPosted: Sat Feb 07, 2015 3:51 pm
by majtejn
Hmm, dunno why. Maybe you could try it from github if you can compile it. By the way it has nothing bad inside, it's only compiled from the source available publicly.
Or just wait a bit as there is a problem with my version that I'll fix as soon as I have time.