Comparison of Backpropagation and Kohonen Self Organising Map (KSOM) Methods in Face Image Recognition

Moonlight, Lady Silk; Faizah, Fiqqih; Suprapto, Yuyun; Pambudiyatno, Nyaris

doi:10.20473/jisebi.7.2.149-161

Cited by 4 publications

(4 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The following steps are taken to achieve the output value: a. input the value of the training data; b. backpropagation of the error value; and c. weight connection adjustment to reduce error value [11,12].…”

Section: Backpropagation Algorithmmentioning

confidence: 99%

See 1 more Smart Citation

Airport Visibility Prediction System to Improve Aviation Safety

Moonlight¹,

Harianto²,

Musadek³

et al. 2023

Proceedings of the International Conference on Advance Transportation, Engineering, and Applied Science (ICATEAS 2022)

Self Cite

View full text Add to dashboard Cite

Security and safety are the most important things in the world of transportation, especially in aviation or air transportation. Safety and security matters are affected by factors such as visibility, wind direction, wind speed, temperature, dew point and barometric pressure. The most important of these factors is visibility. In world aviation, visibility is stated if the distance is more than 5 km. If the visibility is less than 5 km, the flight will be delayed until the visibility is declared safe. Therefore, in this study a system was built to ensure visibility for flight safety and security. The data in this study were taken from the Aviation Meteorology Center of the Meteorology, Climatology and Geophysics Agency, from Juanda International Airport of Surabaya. The visibility system built has an accuracy value of 93.22% or an RMSE value of 0.06776. In this system, the most appropriate number of parameters is determined, namely: the best learning rate is 0.001, the hidden layer unit is 24, and the number of iterations is 10,000 iterations.

show abstract

Section: Backpropagation Algorithmmentioning

confidence: 99%

“…The weight and bias correction between the input layer and the hidden layer are then determined using δ j ( V ij and V 0j respectively). It can be seen in ( 10) (11):…”

Section: A) Backpropagationmentioning

confidence: 99%

Airport Visibility Prediction System to Improve Aviation Safety

Moonlight¹,

Harianto²,

Musadek³

et al. 2023

Proceedings of the International Conference on Advance Transportation, Engineering, and Applied Science (ICATEAS 2022)

Self Cite

View full text Add to dashboard Cite

show abstract

“…Pengurutkan data dilakukann berdasarkan karakteristi dengan kemiripan paling dekat dengan data lainnya akan dikelompokkan menjadi satu klaster. Demikian upaya ini berguna untuk mengarahkan hasil temuan cluster yang sebelumnya tidak diketahui pada data [13], [14].…”

Section: Pendahuluanunclassified

Self Organizing Maps (Kohonen) untuk Cluster Bidang Karya Ilmiah (Skripsi) Mahasiswa Berdasarkan Nilai-Nilai Matakuliah Pendukung Machine Learning

Sunandar¹,

Hasan²

2022

JoSYC

View full text Add to dashboard Cite

The time for the preparation of scientific papers is considered too fast and not suitable for students due to several things in the scope of its preparation. One of the reasons is that the student is unable to complete and the supporting aspects of the field of scientific work being done, many students whose topics of scientific work are in machine learning or the like, but these students are unable to complete their scientific work because they do not understand the theoretical supporting knowledge of the machine learning field. So that it can make the student depressed or even harder to repeat his scientific work the next semester. The scope of machine learning courses are statistics and probability, matrix and linear algebra, algorithms and programming, and data structures. This research was conducted to overcome the problems faced by students, namely knowing the suitability group for the field of student scientific work they will be working on. So that in its preparation students can be responsible in their trial and are of higher quality. The clustering test with the self organizing maps (SOM) algorithm is more stable because the input according to the data owned is not random, only the weighting is done randomly but based on the uniform low (mins) and high (max) limit values. The desired number of clusters is two namely cluster 0 is able to do scientific work based on machine learning and cluster 1 vice versa. The SOM process for 40 student data with a target of two clusters and the results are cluster 0 = 14 students, cluster 1 = 26 students. The result is obtained by increasing the radius = 1, which previously this achievement was not successful if radius = 0.

show abstract

“…The backpropagation method is a systematic method in neural network multiplayer training [15] [16]. Where the output from the network will be compared with the target in order to obtain an output error, then this error is propagated back to improve the network weight in order to minimize errors [17] [18]. Momentum backpropagation processing is used to increase the speed in the recognition process, and in the end it is hoped that it will be able to recognize the Papuan ant nest pattern.…”

Section: Introductionmentioning

confidence: 99%

Image Segmentation Using Haar Wavelet Transform and Artificial Network for Identification of Ant Nest Pattern

Nugraha

2022

JoSYC

View full text Add to dashboard Cite

Ant nest is a powerful medicinal plant to treat various diseases such as heart disease, tumors, cancer, hemorrhoids, tuberculosis, rheumatism, ulcers and prostate disorders. The many benefits generated cause the ant nest to have a selling value in the community, so it is not uncommon to trigger ant nest forgery. This study aims to identify the image of an ant nest by testing the accuracy of the backpropagation network and haar wavelet. Before the identification process with the backpropagation method is carried out, first the process of taking the characteristics of the ant nest is carried out which will later be used as test data input in the identification process. The output of the network will be compared with the target to get the error output. Then these errors are propagated again to increase the network weight to minimize errors. This research resulted in an ant nest identification application. The backpropagation ANN architecture for the training data used is 0.95 momentum, 0.01 target error with MSE criteria, the maximum number of training iterations (epochs) is set at 5000, the learning rate is 0.1 and the number of hidden layers is 20 by using wavelet haar feature extraction level 3 and the backpropagation momentum method can be used. used to identify ant nest images with 100% accuracy.

show abstract

Comparison of Backpropagation and Kohonen Self Organising Map (KSOM) Methods in Face Image Recognition

Cited by 4 publications

References 9 publications

Airport Visibility Prediction System to Improve Aviation Safety

Airport Visibility Prediction System to Improve Aviation Safety

Self Organizing Maps (Kohonen) untuk Cluster Bidang Karya Ilmiah (Skripsi) Mahasiswa Berdasarkan Nilai-Nilai Matakuliah Pendukung Machine Learning

Image Segmentation Using Haar Wavelet Transform and Artificial Network for Identification of Ant Nest Pattern

Contact Info

Product

Resources

About