Profile MMSE - pomiar 1

• Dane: statystyki opisowe (6 cech) + charakteryzacja kształtów roskładu profili (4 cechy: średnia, odchylenie, skośność i kurtoza z gęstości)

Płeć

• Test: GM stand 10x5CV
• Nic powyżej baserate (lub chociaż 30% )

Klasyfikacja 2 klasowa

• KLasy: przeskalowane wyniki testu < 11.3
• Testy: GM + std + 10x5CV
  • 58.1%/12.1 SVN gauss c=1.0, bias=0.5
  • 55.7%/17.1 kNN k=1 euclid
  • Reszta ponizej baserate

Wizualizacja QPC

>> w=qpctrain(xn,c2,'display','short')
procedure     = qpc
dataname      = data
vectors       = 43
features      = 10
OptMethod     = gd
learningRate  = 0.100000
prototypes    = 0
proto lr      = 0.000000
eps           = 0.001000
start init.   = 10
end init      = 3
multistart    = yes
maxIterations = 1000
qpc function  = @(wx,p)qpcfunction(x,ci,wx,[],func)
function      = @(x)f_gauss(x,parameters.beta)
beta          = 0.100000
Performing 10 initiations - Gradient based optimization (MultiStart)
Sprawdzic dokladniej czy dobrze
6 initiations was killed
Finish N = 220   QPC = 0.088294 [3 ]
   -0.2625   -0.2240   -0.5884   -0.4564   -0.0720   -0.3458   -0.2978   -0.1616    0.2821   -0.0852

Finish N = 280   QPC = 0.081809 [2 ]
   -0.2626   -0.2302    0.3327   -0.1507   -0.2403   -0.4847   -0.5708   -0.2200    0.2416   -0.1394

Finish N = 295   QPC = 0.082032 [7 ]
   -0.3728   -0.2509    0.3371   -0.1551   -0.2610   -0.3071   -0.5174   -0.2834    0.2985   -0.2467

Finish N = 315   QPC = 0.088345 [8 ]
   -0.2133   -0.2457   -0.5858   -0.4604   -0.1122   -0.3483   -0.3022   -0.1561    0.2857   -0.0886

Best:   0.088345     -0.2133  -0.2457  -0.5858  -0.4604  -0.1122  -0.3483  -0.3022  -0.1561  0.2857  -0.0886
---------------Sumary--------------------------
Initialization  8 was the best
N     QPC      W
0  0.0883      -0.2133  -0.2457  -0.5858  -0.4604  -0.1122  -0.3483  -0.3022  -0.1561  0.2857  -0.0886
-----------------------------------------------
procedure     = qpc
dataname      = data
vectors       = 43
features      = 10
OptMethod     = gd
learningRate  = 0.100000
prototypes    = 0
proto lr      = 0.000000
eps           = 0.001000
start init.   = 10
end init      = 3
multistart    = yes
maxIterations = 1000
qpc function  = @(wx,p)qpcfunction(x,ci,wx,[],func)
function      = @(x)f_gauss(x,parameters.beta)
beta          = 0.100000
Performing 10 initiations - Gradient based optimization (MultiStart)
Sprawdzic dokladniej czy dobrze
6 initiations was killed
Finish N = 325   QPC = 0.060643 [8 ]
   -0.2338   -0.3198    0.3677   -0.4303   -0.0890   -0.3894    0.0253   -0.2643   -0.5001   -0.2063

Finish N = 350   QPC = 0.060591 [7 ]
   -0.1145   -0.3625    0.3456   -0.4490   -0.1763   -0.4404    0.0713   -0.2678   -0.4229   -0.2319

Finish N = 385   QPC = 0.060768 [3 ]
   -0.1647   -0.2409    0.3231   -0.4437   -0.0999   -0.5500   -0.0499   -0.2346   -0.4657   -0.1634

Finish N = 430   QPC = 0.060598 [1 ]
   -0.1338   -0.2715    0.3599   -0.4280   -0.1425   -0.4570    0.0535   -0.2980   -0.4398   -0.2852

Best:   0.060768     -0.1647  -0.2409  0.3231  -0.4437  -0.0999  -0.5500  -0.0499  -0.2346  -0.4657  -0.1634
---------------Sumary--------------------------
Initialization  3 was the best
N     QPC      W
0  0.0608      -0.1647  -0.2409  0.3231  -0.4437  -0.0999  -0.5500  -0.0499  -0.2346  -0.4657  -0.1634
-----------------------------------------------