Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Badger Storm]

Will the 40.10 version work okay with 40.14?

[lethosor]

Probably not. Which platform are you using?

[sv-esk]

Will the 40.10 version work okay with 40.14?

Try to copy v040.14.ini file from "operating system name" folder from  https://github.com/splintermind/Dwarf-Therapist/tree/DF2014/share/memory_layouts
to      \Dwarf Therapist\share\memory_layouts\"operating system name"\    folder on your computer.
If not working, or you want v28 (see patch notes https://github.com/splintermind/Dwarf-Therapist/wiki/Change-Log) ask here somebody to compile v28 for your OS.
Or try yourself https://github.com/splintermind/Dwarf-Therapist/blob/DF2014/BUILDING.rst
(if already compiled for osx and linux why no links on first post?)

------------------------
dwarf-therapist v28 for 0.40.14 linux(*.deb) i386    qt5
http://www.mediafire.com/download/x75jomdr7jjlwar/dwarf-therapist_28.0-1_i386.deb

[thistleknot]

[Geometric Mean]
...
  A^Wa * B^Wb * C^Wc * D^Wd

where Wa + Wb + Wc + Wd = 1. Though I suspect this won't be what you actually need, either.

Maklak recommending something similar.

(a^wa * b^wb * c^wc)^(-3*(wa+wb+wc))

Spoiler (click to show/hide)

In the document I was using, it had 3 values and 3 weights, so I'm guessing that's where the -3 comes in.

However... the results were pretty weird.  They didn't seem to correlate with higher numbers = better matches

I did try

(a^wa * b^wb * c^wc)^((wa+wb+wc))

which seemed a bit more promising, does anyone know how to apply a geometric mean to 3 variables in a decision that each have their own weights?  What if I want to extend it up to 4 variables, or 5 variables?

In a weighted average I divide by the sum of the weights, what about in a geometric weighted mean?  Is it to the power of the sum of weights?

I don't know how to implement a geometric mean.  My math never covered it, the concept is foreign to me.  I've read about it, it seems to be deriving a locally derived center amongst a multi dimensional object.  But I'm at a loss for how to implement any meaningful method using it.

Update
I think I got what I wanted

thanks to this doc http://bit.csc.lsu.edu/trianta/EditedBook_CHAPTERS/EEEE1.pdf

and others reminding me that dividing by the numbers I factor against cancel each other out... I need to raise the value by a power equal to it's weight, and factor these values against each other.

But I did have to modify the formula

(a^weight)*(b^weight)*(c^weight)
drop any dividing, gives me a rankable list to derive %'s from.

http://dffd.wimbli.com/file.php?id=10014

[lethosor]

(if already compiled for osx and linux why no links on first post?)

[OSX] Dwarf Therapist 28.pre-release | DF 40.14

Splinterz, could you update the link in the first post?
Edit: Thanks!

[thistleknot]

I don't wish to keep spamming this thread, but... DT has a secondary feature known as a Multi-Criteria Decision Analysis system.  That's the Role % that is derived and used in the Labor Optimizer.

the larger conversation I'm hoping to have here is...

I'm proposing changes to how that operates by introducing a

Weighted Product method.

alongside a

Weighted Average method

Then averaging the relevant ranks [represented as %'s] of the weighted product method along with the ~[relevant ranks of the] weighted average method to produce a new final output %.  so [a + b] / 2 = c

I'm trying to have that conversation with the spreadsheet document I posted.

I don't know if combining two MCDA methods into one system is a good idea, maybe it is.  Any objections, does anyone care?  Is Weighted Average good enough?

Update

Let us assume that we evaluate solutions in a specific problem situation using several criteria. Let us further assume that more is better in each criterion. Then, among all possible solutions, we are ideally interested in those solutions that perform well in all considered criteria. However, it is unlikely to have a single solution that performs well in all considered criteria. Typically, some solutions perform well in some criteria and some perform well in others. Finding a way of trading off between criteria is one of the main endeavors in the MCDM literature.
Mathematically, the MCDM problem corresponding to the above arguments can be represented as

        "max" q

        subject to

            q ∈ Q

I'm hoping to expand q, but I'm not sure if I'm going about it the correct way, is averaging these two comparative methods a good idea?

[Chimerat]

Just in case I missed a feature somewhere (because DT has a lot ), is there any way to check relationships between Dwarves? For examples, my Expedition Leader and Doctor are apparently holding mutual grudges, and none of my other Dwarves are more than "Friend" or "Friendly Terms" with each other.

But I didn't know about the grudge until I was individually checking my seven to see if anyone was getting interested in another Dwarf.

[thistleknot]

so I checked out geometric mean and was comparing it and seeing how different values/weights affect the output

However. I forgot a very important step in weighted models...

Raising #'s to their nth power gets tricky with weights when the power switches from fractions to whole numbers.

I found the behavior when .3 ^ by .25 or 1.5 weight.

.3 gets BIGGER when raised by .25 but smaller when raised by 1.5

and numbers above 1 get the opposite treatment.

ex .3 ^ .25 = .74
ex .3 ^ 1.5 = .16

ex 2 ^ .25 = 1.18
ex 2 ^ 1.5 = 2.82

I just realized that raising #'s in a weighted product model requires one to set the weights to proportions...

the excel formula I have is this:
((a^weighta)*(b^weightb)*(c^weightc))^(1/(count(weights)))

Supposedly the #'s don't need to be normalized against each other, but the weights need to (i.e. sum of weights = 1).

[splinterz]

Just in case I missed a feature somewhere (because DT has a lot ), is there any way to check relationships between Dwarves? For examples, my Expedition Leader and Doctor are apparently holding mutual grudges, and none of my other Dwarves are more than "Friend" or "Friendly Terms" with each other.

But I didn't know about the grudge until I was individually checking my seven to see if anyone was getting interested in another Dwarf.

no, there's currently no way to view the relationships between dwarves.

[Chimerat]

no, there's currently no way to view the relationships between dwarves.

Okay. Thanks for confirming that. Sometimes I miss features because I don't know where to go looking.

[lxnt]

Is everyone here running debug builds? Because I get bitten by

Code: [Select]

QSpiAccessible::accessibleEvent not handled:  "2"  obj:  QMessageBox(0x1553bd0) ""
FIXME: handle dialog start.

every time I rebuild the therapist and forget the -f Makefile.Debug part.

System's xubuntu 14.04.1

[Hello71]

Is everyone here running debug builds? Because I get bitten by

Code: [Select]

QSpiAccessible::accessibleEvent not handled:  "2"  obj:  QMessageBox(0x1553bd0) ""
FIXME: handle dialog start.

every time I rebuild the therapist and forget the -f Makefile.Debug part.

System's xubuntu 14.04.1

1. make debug
2. debug builds are automatically done through travis ci; I also use them by default when developing. however, I haven't had the time yet to test 28.0.

[thistleknot]

Spoiler: context (click to show/hide)

The old sheet here: http://dffd.wimbli.com/file.php?id=10014

was set to 10 generated dwarfs, and the weighted average ~= the weighted geometric mean results.

However... since the values are arbitrarily pre transformed before compared with each other, I thought it would be more "pure" to use the "raw" values.

So...

I redid all the roles using raw values vs transformed values as inputs for a weighted geometric mean.  I dropped the file size down by a 3rd

and cleaned up a lot of the sheets.

http://dffd.wimbli.com/file.php?id=10023

Please, if you have time, I'd love anyone to see if I f'd it up.

It currently generates 100 Monte Carlo dwarf's with varying skills, traits, and attributes.

It appears to produce a comparative relationship.

The first 4 sheets are the important "results"
sheet 1 = global variables and the final generated roles
sheet 2 - 4 = 3 pre-defined roles (They are editable, so one can make any role of up to 5 input vars) and how the raw values are fed into them to produce a weighted geometric mean.

**Update**
Much easier way to view the way the changes affect things
http://imgur.com/IuLdwjQ

I think the reason it works is because it's like finding the volume of a 3d area (but not limited to just 3d) (height * width * length), but finding a volume's center I guess.  When all the values are combined [using a WGM role setup] into single role matrix; all the dimensions comparable statistics like min, max, mean, median, etc, are built into the 1 dimensional output: the role matrix.

**Update**
I updated the setup to include virtual preferences.

[lxnt]

1. make debug
2. debug builds are automatically done through travis ci; I also use them by default when developing. however, I haven't had the time yet to test 28.0.

Yeah, make debug. Whatever.

The question was - is the release build known to be broken?

[Chimerat]

I got to play a few hours of DF over the last few days and I have to say that I love the ability to set animals to be butchered from DT itself. Much easier than trying to scroll the menus in game.

EDIT: And now I see that there's a new update up today. Cool. Trading tweaks. (Though I'd still love a way to just search for something in the caravan. *sighs*)