QPC + LVQ1 - Test 10CV

Dataset			Model
	vec	feat	acc.	std.	#P	std.	#K	std.
Appendicities	106	7	86.09	8.45	1.00	0.00	2.50	1.20
Australian	690	14	86.08	4.14	1.00	0.00	2.00	0.00
Bitsymetry15	100	15	82.00	11.66	1.00	0.00	3.00	0.00
Bitsymetry20	100	20	81.00	7.00	1.00	0.00	3.00	0.00
Breast Cancer W. (nomis)	683	9	96.20	1.75	1.00	0.00	2.00	0.00
Czerniak (trs)	250	14	85.67	6.22	1.00	0.00	4.00	0.00
Glass	214	9	60.37	10.34	1.10	0.30	4.50	0.92
Heart	270	13	78.89	8.45	1.00	0.00	2.00	0.00
Ionosphere	200	34	78.44	7.97	1.00	0.00	3.10	0.70
Iris	150	4	96.00	4.42	1.00	0.00	3.00	0.00
L. Breast (nomis)	277	9	72.55	4.38	1.00	0.00	2.00	0.00
Led500	500	7	58.45	8.05	1.10	0.30	9.60	0.49
Parity-10	1024	10	96.09	3.85	1.00	0.00	6.80	0.40
Parity-8	256	8	94.52	4.97	1.00	0.00	6.30	0.64
Voting	435	16	95.15	3.50	1.00	0.00	2.00	0.00
Wine	178	13	96.04	5.67	1.90	0.30	3.10	0.30

Network parameters

 Method            lvq+qpc
 Initialization    5
 LVQ learnign rate 0.010000
 Attraction force  0.050000
 Max. projections  5
 K range           2 10 20 
 Attraction step   100
 Precision (eps)   0.030000

QPC parameters (default)

 
 lrate    = 0.1;       % learning rate
 maxiter  = 1000;      % MAX number of iterations
 init     = 5;         % number of initializations
 function = gauss;     %
 width    = 0.1;       % function width
 eps      = 0.001;     %

Dataset			QPC/LVQ1						SVM		MLP		1-NN
	vec	feat	acc.	std.	#P	std.	# K	std.	acc.	std.	acc.	std.	acc.	std.
Appendicities	106	7	85.40	1.51	1.00	0.00	2.74	0.49
Australian	690	14	84.86	0.43	1.00	0.00	2.00	0.00
Brest	683	9
Breast (Wisconsin)	683	9
Czerniak-trs	250	14
Glass	214	9
Heart	270	13
Ionosphere	200	34
Iris	150	4

— Marek 2009/03/17 15:58

QPC + LVQ1 - Test 10CV

Ustawienia

10x10CV