Adaptive neural network classifier for extracted invariants of handwriten digits
We propose an adaptive activation function of neural network classifier for isolated handwritten digits that undergo basic transformations. The utilized network is a backpropagation network with sigmoid and arctangent activation functions. The performance of network with both activation functions is...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Universiti Utara Malaysia
2004
|
| Subjects: | |
| Online Access: | https://repo.uum.edu.my/id/eprint/1039/1/L._H._kENG.pdf |
| _version_ | 1825739758046806016 |
|---|---|
| author | Keng, L. H. Shamsuddin, Siti Mariyam |
| author_facet | Keng, L. H. Shamsuddin, Siti Mariyam |
| author_sort | Keng, L. H. |
| collection | UUM |
| description | We propose an adaptive activation function of neural network classifier for isolated handwritten digits that undergo basic transformations. The utilized network is a backpropagation network with sigmoid and arctangent activation functions. The performance of network with both activation functions is compared. The results show that the network applying an adaptive activation function between layers converged much faster compared to non-adaptive activation functions with 50% iterations reduction. In this study, we also present experimental results of feature extraction between Zernike and d-geometric for better feature representations. Results show that Zernike features are better at representing isolated handwritten digits compared to d-geometric features with an accuracy up to 87%.
|
| first_indexed | 2024-07-04T05:14:56Z |
| format | Article |
| id | uum-1039 |
| institution | Universiti Utara Malaysia |
| language | English |
| last_indexed | 2024-07-04T05:14:56Z |
| publishDate | 2004 |
| publisher | Universiti Utara Malaysia |
| record_format | eprints |
| spelling | uum-10392010-09-05T06:53:57Z https://repo.uum.edu.my/id/eprint/1039/ Adaptive neural network classifier for extracted invariants of handwriten digits Keng, L. H. Shamsuddin, Siti Mariyam QA75 Electronic computers. Computer science We propose an adaptive activation function of neural network classifier for isolated handwritten digits that undergo basic transformations. The utilized network is a backpropagation network with sigmoid and arctangent activation functions. The performance of network with both activation functions is compared. The results show that the network applying an adaptive activation function between layers converged much faster compared to non-adaptive activation functions with 50% iterations reduction. In this study, we also present experimental results of feature extraction between Zernike and d-geometric for better feature representations. Results show that Zernike features are better at representing isolated handwritten digits compared to d-geometric features with an accuracy up to 87%. Universiti Utara Malaysia 2004 Article PeerReviewed application/pdf en https://repo.uum.edu.my/id/eprint/1039/1/L._H._kENG.pdf Keng, L. H. and Shamsuddin, Siti Mariyam (2004) Adaptive neural network classifier for extracted invariants of handwriten digits. Journal of ICT, 3 (1). pp. 1-17. ISSN 1675-414X http://jict.uum.edu.my |
| spellingShingle | QA75 Electronic computers. Computer science Keng, L. H. Shamsuddin, Siti Mariyam Adaptive neural network classifier for extracted invariants of handwriten digits |
| title | Adaptive neural network classifier for extracted invariants of handwriten digits |
| title_full | Adaptive neural network classifier for extracted invariants of handwriten digits |
| title_fullStr | Adaptive neural network classifier for extracted invariants of handwriten digits |
| title_full_unstemmed | Adaptive neural network classifier for extracted invariants of handwriten digits |
| title_short | Adaptive neural network classifier for extracted invariants of handwriten digits |
| title_sort | adaptive neural network classifier for extracted invariants of handwriten digits |
| topic | QA75 Electronic computers. Computer science |
| url | https://repo.uum.edu.my/id/eprint/1039/1/L._H._kENG.pdf |
| work_keys_str_mv | AT kenglh adaptiveneuralnetworkclassifierforextractedinvariantsofhandwritendigits AT shamsuddinsitimariyam adaptiveneuralnetworkclassifierforextractedinvariantsofhandwritendigits |