Flotation Performance Recognition Based on Dual-modality Convolutional Neural Network Adaptive Transfer Learning

Yi-peng LIAO; Jie-jie YANG; Zhi-gang WANG; Wei-xing WANG

doi:10.3788/gzxb20204910.1015001

Journals >Acta Photonica Sinica >Volume 49 >Issue 10 >Page 1015001 > Article

Acta Photonica Sinica
Vol. 49, Issue 10, 1015001 (2020)

Flotation Performance Recognition Based on Dual-modality Convolutional Neural Network Adaptive Transfer Learning

Yi-peng LIAO¹, Jie-jie YANG¹, Zhi-gang WANG², and Wei-xing WANG^1、*

Author Affiliations

¹College of Physics and Information Engineering，Fuzhou University，Fuzhou 350108，China

²Fujian Jindong Mining Co. Ltd. ，Sanming，Fujian 365101，China

show less

DOI: 10.3788/gzxb20204910.1015001 Cite this Article

Yi-peng LIAO, Jie-jie YANG, Zhi-gang WANG, Wei-xing WANG. Flotation Performance Recognition Based on Dual-modality Convolutional Neural Network Adaptive Transfer Learning[J]. Acta Photonica Sinica, 2020, 49(10): 1015001 Copy Citation Text

show less

Fig. 1. Flotation foam dual-modality image

Download full size | View in the Article

Fig. 2. Dual-modality CNN feature extraction and recognition model

Download full size | View in the Article

Fig. 3. Double hidden layer autoencoder extreme learning machine

Download full size | View in the Article

Fig. 4. Adaptive transfer learning model

Download full size | View in the Article

Fig. 5. The operation effect of each optimization algorithm

Download full size | View in the Article

Fig. 6. Performance test of double hidden layer autoencoder extreme learning machine

Download full size | View in the Article

Fig. 7. Test results of different training samples

Download full size | View in the Article

Fig. 8. Comparison of performance recognition results

Download full size | View in the Article

Function	Function formula	Parameter range	Optimum value
Rosenbrock	$f_{1} (x) = \sum_{i = 1}^{N - 1} [100 (x_{i + 1} - x_{i}^{2})^{2} + (x_{i} {- 1)}^{2}]$	[-10，10]	0
Rastrigrin	$f_{2} (x) = \sum_{i = 1}^{N} [x_{i}^{2} - 10 c o s (2 π x_{i}^{2}) + 10]$	[-100，100]	0
Schewefel	$f_{3} (x) = 418.9829 N + [- x_{i} s i n (\sqrt[]{x_{i}})]$	[-500，500]	0
Shubert	$f_{4} (x) = 0.5 - \frac{s i n^{2} \sqrt[]{x_{1}^{2} + x_{2}^{2}} - 0.5}{[1 + 0.001 (x_{1}^{2} + x_{2}^{2} {)]}^{2}}$	[-100，100]	1
Schaffer	$f_{5} (x) = \{\sum_{i = 1}^{5} i c o s [(i + 1) x + i]\} \{\sum_{i = 1}^{5} i c o s [(i + 1) y + i]\}$	[-100，100]	-186.730 9

Table 1. Benchmark functions information

View in the Article

$λ$ $Δ θ$	0.1π	0.2π	0.3π	0.4π	0.5π
1.1	97.89/205.1	98.56/210.2	99.09/196.5	98.79/213.6	97.37/212.4
1.2	97.88/208.8	98.84/204.4	99.20/204.7	99.16/204.6	98.31/198.5
1.3	98.14/199.4	98.78/199.4	99.15/202.6	99.35/198.4	97.81/201.6
1.4	98.56/197.6	98.95/196.4	99.64/198.7	99.64/202.7	98.33/208.1
1.5	98.85/204.5	99.35/196.6	99.70/192.9	99.81/205.3	99.11/205.3
1.6	99.36/209.4	99.12/201.3	99.95/190.2	99.85/203.8	98.76/197.4
1.7	99.87/201.8	99.76/195.2	99.92/191.8	99.65/206.2	99.23/196.6
1.8	99.78/194.3	99.23/198.8	99.63/194.3	99.54/199.5	99.15/196.1
1.9	99.46/199.2	98.84/208.3	99.24/201.8	99.21/198.1	98.86/195.8
2.0	98.95/206.4	98.97/207.1	99.26/202.7	98.89/201.3	98.15/202.3

Table 2. The average optimal value probability (%) and the number of iterations

View in the Article

Data set	Lonosphere		Shuttle		USPS
Test item	Accuracy/%	Times/s	Accuracy/%	Times/s	Accuracy/%	Times/s
KELM	90.12	1.253	91.31	26.930	91.75	162.961
AE⁃KELM	94.23	2.014	95.52	38.237	95.81	720.834
DAE⁃KELM	96.45	4.587	98.31	45.417	98.35	1 646.250

Table 3. Test results of data sets for various KELM algorithm

View in the Article

Algorithm	Feature extraction method	Classification algorithm	Accuracy/%	Standard deviation/%
Ref. [6]	Alexnet	Random forest	85.93	4.24
Ref. [7]	CNN feature statistics	PW mean⁃shift	91.25	1.82
Ref. [8]	Two layers CNN	SVM	89.06	2.95
Ref. [9]	Alexnet transfer learning	Random forest	93.02	2.68
Proposed	Dual⁃modality alexnet transfer learning	Adaptive DAE⁃KELM	96.83	1.96

Table 4. Recognition effect of different methods

Download Citation

Save the article for my favorites

Paper Information