Page 65 - Fister jr., Iztok, Andrej Brodnik, Matjaž Krnc and Iztok Fister (eds.). StuCoSReC. Proceedings of the 2019 6th Student Computer Science Research Conference. Koper: University of Primorska Press, 2019
P. 65
224
112
56
28
14
4096
4096
K
25088 1
Fully
224 11 Connected 3
512 512 512 Fully Fully + Softmax
Block 5 ConnectedC1onnected 2
1
256 256 256 512 512 512 Flatten
Block 3 Block 4
6464 128128
Block 1 Block 2
Figure 3: The architecture of the VGG16 convolutional neural network.
5. RESULTS Metrics Baseline GWOTLT
The obtained performance results from the conducted ex- Time [s] 49.10 ± 1.85 759.10 ± 59.67
periments are summarized in Table 3. Focusing on the time AUC [%] 87.00 ± 9.19 91.00 ± 7.75
metrics, the reported results are expected, with the lowest F − 1 [%] 86.27 ± 11.03 91.45 ± 6.81
time complexity being achieved by the Baseline method. On Precision [%] 88.62 ± 10.37 90.89 ± 11.36
the other side, the proposed GWOTLT method is expected Recall [%] 86.00 ± 17.13 93.00 ± 6.75
to have a higher time complexity in general due to the it-
erative nature of the proposed method. In our case, the Kappa 0.74 ± 0.18 0.82 ± 0.15
GWOTLT method performed worse in the aspect of time
complexity, roughly by a factor 15. Table 3: Comparison of average times, accuracies,
AUCs, F − 1 scores, precisions, recalls and kappa
Analyzing presented classification performance metrics, the coefficients with standard deviations over 10-fold
GWOTLT method is standing out with achieved best results cross-validation.
on all of the reported performance metrics. The AUC, F −1,
precision and recall metrics are higher by a margin of 4%, In the future, we would like to expand our work to include
5.18%, 2.27%, 7% respectively in comparison to the baseline various CNN architectures as a convolutional base for our
method. Focusing on the kappa coefficient values, we can ob- GWOTLT method and also evaluate the performance of the
serve that the GWOTLT achieved a near-perfect agreement proposed method against various medical imaging datasets.
with kappa coefficient at 0.82 and outperformed the base-
line method by a margin of 0.08. Looking at the standard Acknowledgments
deviations of the reported classification average metric val-
ues, we can observe that for all classification metrics, except The authors acknowledge the financial support from the
for the precision, the best performing GWOTLT method is Slovenian Research Agency (Research Core Funding No. P2-
showing the smallest standard deviation. The greatest im- 0057).
provement of lowering the standard deviation the GWOTLT
achieved for the recall metric by a margin of 10.38%, while 7. REFERENCES
the worst standard deviation is obtained for the precision
metric where the GWOTLT lacks behind just by 0.99%. [1] S. U. Akram, J. Kannala, L. Eklund, and J. Heikkil¨a.
Cell segmentation proposal network for microscopy
6. CONCLUSIONS image analysis. In Deep Learning and Data Labeling
for Medical Applications, pages 21–29. Springer, 2016.
In this paper, we presented the GWOTLT method which is
a nature-inspired, population-based metaheuristics method [2] E. Al Hadhrami, M. Al Mufti, B. Taha, and
for tuning the transfer learning approach of training the deep N. Werghi. Transfer learning with convolutional neural
CNN. The GWOTLT method was implemented utilizing the networks for moving target classification with
GWO optimization algorithm and applied to the problem of micro-doppler radar spectrograms. In 2018
identification of brain hemorrhage from the head CT scan International Conference on Artificial Intelligence and
images. The results obtained from the conducted exper- Big Data (ICAIBD), pages 148–154. IEEE, 2018.
iments have proven that the proposed GWOTLT method
seems to be very promising for the task of transfer learning [3] U. Balasooriya and M. Perera. Intelligent brain
tuning achieving higher classification performance for all of hemorrhage diagnosis system. In 2011 IEEE
the measured classification metrics. International Symposium on IT in Medicine and
Education, volume 2, pages 366–370. IEEE, 2011.
[4] P. Chang, E. Kuoy, J. Grinband, B. Weinberg,
M. Thompson, R. Homo, J. Chen, H. Abcede,
StuCoSReC Proceedings of the 2019 6th Student Computer Science Research Conference 65
Koper, Slovenia, 10 October
112
56
28
14
4096
4096
K
25088 1
Fully
224 11 Connected 3
512 512 512 Fully Fully + Softmax
Block 5 ConnectedC1onnected 2
1
256 256 256 512 512 512 Flatten
Block 3 Block 4
6464 128128
Block 1 Block 2
Figure 3: The architecture of the VGG16 convolutional neural network.
5. RESULTS Metrics Baseline GWOTLT
The obtained performance results from the conducted ex- Time [s] 49.10 ± 1.85 759.10 ± 59.67
periments are summarized in Table 3. Focusing on the time AUC [%] 87.00 ± 9.19 91.00 ± 7.75
metrics, the reported results are expected, with the lowest F − 1 [%] 86.27 ± 11.03 91.45 ± 6.81
time complexity being achieved by the Baseline method. On Precision [%] 88.62 ± 10.37 90.89 ± 11.36
the other side, the proposed GWOTLT method is expected Recall [%] 86.00 ± 17.13 93.00 ± 6.75
to have a higher time complexity in general due to the it-
erative nature of the proposed method. In our case, the Kappa 0.74 ± 0.18 0.82 ± 0.15
GWOTLT method performed worse in the aspect of time
complexity, roughly by a factor 15. Table 3: Comparison of average times, accuracies,
AUCs, F − 1 scores, precisions, recalls and kappa
Analyzing presented classification performance metrics, the coefficients with standard deviations over 10-fold
GWOTLT method is standing out with achieved best results cross-validation.
on all of the reported performance metrics. The AUC, F −1,
precision and recall metrics are higher by a margin of 4%, In the future, we would like to expand our work to include
5.18%, 2.27%, 7% respectively in comparison to the baseline various CNN architectures as a convolutional base for our
method. Focusing on the kappa coefficient values, we can ob- GWOTLT method and also evaluate the performance of the
serve that the GWOTLT achieved a near-perfect agreement proposed method against various medical imaging datasets.
with kappa coefficient at 0.82 and outperformed the base-
line method by a margin of 0.08. Looking at the standard Acknowledgments
deviations of the reported classification average metric val-
ues, we can observe that for all classification metrics, except The authors acknowledge the financial support from the
for the precision, the best performing GWOTLT method is Slovenian Research Agency (Research Core Funding No. P2-
showing the smallest standard deviation. The greatest im- 0057).
provement of lowering the standard deviation the GWOTLT
achieved for the recall metric by a margin of 10.38%, while 7. REFERENCES
the worst standard deviation is obtained for the precision
metric where the GWOTLT lacks behind just by 0.99%. [1] S. U. Akram, J. Kannala, L. Eklund, and J. Heikkil¨a.
Cell segmentation proposal network for microscopy
6. CONCLUSIONS image analysis. In Deep Learning and Data Labeling
for Medical Applications, pages 21–29. Springer, 2016.
In this paper, we presented the GWOTLT method which is
a nature-inspired, population-based metaheuristics method [2] E. Al Hadhrami, M. Al Mufti, B. Taha, and
for tuning the transfer learning approach of training the deep N. Werghi. Transfer learning with convolutional neural
CNN. The GWOTLT method was implemented utilizing the networks for moving target classification with
GWO optimization algorithm and applied to the problem of micro-doppler radar spectrograms. In 2018
identification of brain hemorrhage from the head CT scan International Conference on Artificial Intelligence and
images. The results obtained from the conducted exper- Big Data (ICAIBD), pages 148–154. IEEE, 2018.
iments have proven that the proposed GWOTLT method
seems to be very promising for the task of transfer learning [3] U. Balasooriya and M. Perera. Intelligent brain
tuning achieving higher classification performance for all of hemorrhage diagnosis system. In 2011 IEEE
the measured classification metrics. International Symposium on IT in Medicine and
Education, volume 2, pages 366–370. IEEE, 2011.
[4] P. Chang, E. Kuoy, J. Grinband, B. Weinberg,
M. Thompson, R. Homo, J. Chen, H. Abcede,
StuCoSReC Proceedings of the 2019 6th Student Computer Science Research Conference 65
Koper, Slovenia, 10 October
