Skip to main content

Table 2 The average accuracy of and time needed by EM algorithm, Q and C neural networks with different numbers of neurons in 2D and 3D spaces with \(K=5\) and 8 Gaussian clusters. The notation \((\cdot )\) stands for the architecture of a neural network as described in Eq. 15

From: Superiority of quadratic over conventional neural networks for classification of gaussian mixture data

 

Accuracy (%)

Time (s)

Accuracy (%)

Time (s)

 

D = 2, K = 5

D = 3, K = 5

C(2-3)

\(92.36 \pm 7.89\)

\(11.5 \pm 5.0\)

\(84.72 \pm 10.32\)

\(13.7 \pm 5.8\)

C(2-10-3)

\(95.36 \pm 5.66\)

\(23.3 \pm 14.1\)

\(92.15 \pm 8.04\)

\(24.8 \pm 8.6\)

C(2-100-3)

\(95.47 \pm 5.68\)

\(62.7 \pm 27.9\)

\(92.54 \pm 8.01\)

\(71.8 \pm 29.2\)

Q(2-3)

\(95.53 \pm 5.66\)

\(15.8 \pm 6.7\)

\(92.74 \pm 7.98\)

\(17.6 \pm 7.4\)

EM

\(95.60 \pm 5.65\)

 

\(92.90 \pm 7.97\)

 
 

D = 2, K = 8

D = 3, K = 8

C(2-3)

\(83.84 \pm 5.53\)

\(22.8 \pm 8.0\)

\(76.47 \pm 6.31\)

\(28.8 \pm 10.6\)

C(2-10-3)

\(87.90 \pm 3.75\)

\(47.6 \pm 17.1\)

\(82.45 \pm 5.36\)

\(62.6 \pm 18.5\)

C(2-100-3)

\(88.06 \pm 3.70\)

\(122.6 \pm 39.2\)

\(82.68 \pm 5.34\)

\(151.8 \pm 41.4\)

Q(2-3)

\(88.13 \pm 3.67\)

\(33.3 \pm 11.0\)

\(82.79 \pm 5.31\)

\(46.2 \pm 13.3\)

EM

\(88.19 \pm 3.64\)

 

\(82.86 \pm 5.32\)

Â