Applied Computational Intelligence and Soft Computing / 2012 / Article / Tab 1 / Research Article
Nonnegative Matrix Factorizations Performing Object Detection and Localization Table 1 Algorithm performances when applied to CarData, USPS, and ORL dataset, respectively. Reported values refer to the lowest and highest values of the factor rank
π
as previously described.
ββββββββββCarData Rank 20 110 Method MSE Time
o
r
t
h
(
π
)
MSE Time
o
r
t
h
(
π
)
NMF
2
.
4
4
1
π
9
275
8
.
7
4
1
1
π
4
1
.
4
5
7
π
9
453
4
.
4
4
3
5
π
5
LNMF
2
.
4
0
4
π
1
0
292 4.9734
2
.
3
7
3
π
1
0
472 10.2373 NMFsc
2
.
5
5
9
π
9
695
6
.
7
8
1
8
π
9
1
.
4
2
2
π
9
1265
1
.
5
8
2
5
π
9
DLPP
2
.
6
6
4
π
9
2271 1.5627
1
.
6
5
7
π
9
2591 3.3221 ββββββββββUSPS Rank 80 220 Method MSE Time
o
r
t
h
(
π
)
MSE Time
o
r
t
h
(
π
)
NMF
1
.
2
9
7
π
4
397
2
.
8
1
6
6
π
4
3
.
0
3
1
π
3
847
1
.
2
1
4
2
π
5
LNMF
1
.
3
3
1
π
5
374 6.6387
1
.
6
0
9
π
4
1427 6.4695 NMFsc
1
.
3
1
8
π
4
777
5
.
2
8
5
4
π
4
5
.
5
6
8
π
3
1409
2
.
7
7
6
1
π
4
DLPP
1
.
5
0
7
π
4
637 3.4077
1
.
2
4
9
π
3
1144 3.2623 ββββββββββORL Rank 20 80 Method MSE Time
o
r
t
h
(
π
)
MSE Time
o
r
t
h
(
π
)
NMF
1
.
0
2
7
π
9
496
1
.
5
7
0
1
π
5
5
.
4
1
3
π
8
705
6
.
0
5
7
7
π
5
LNMF
3
.
1
0
4
π
1
0
556 4.4656
3
.
0
8
0
π
1
0
781 8.8920 NMFsc
1
.
4
2
5
π
9
1362
1
.
0
7
6
2
π
1
0
6
.
1
8
3
π
8
2164
2
.
2
6
7
4
π
9
DLPP
1
.
3
2
3
π
9
14824 1.7690
8
.
1
4
5
π
8
15278 3.4647
