摘要(英) |
Image recognition is popular in artificial intelligence and can be applied to many fields, such as handwritten digit recognition, license plate recognition, face recognition, object recognition and so on. Using deep learning methods can effectively extract features and reduce costs. But, creating a good classification model requires consideration of many factors. For example: the appropriate model architecture, the appropriate optimization method, the appropriate parameter settings, and so on.
The butterfly images of this experiment are taken from ImageNet, and the butterfly identification models are constructed by the convolutional neural network. Several factors that may affect the butterfly identification model are selected as the objects of discussion and comparison. It is observed from the experimental results that the size of the dropout ratio, the size and placement of the pooling layer, the different optimization algorithms and the different layers of convolution layers all affect the ability of the butterfly identification model. Therefore, when constructing the model, these factors must be carefully chosen, and their influence on the model cannot be ignored. |
參考文獻 |
[1].黃安埠 (2017)。深入淺出深度學習-原理剖析與Python實踐。電子工業出版社。
[2].鄭澤宇、顧思宇 (2017)。Tensorflow實戰Google深度學習框架。電子工業出版社。
[3].林大貴 (2017)。TensorFlow + Keras深度學習人工智慧實務應用。博碩出版社。
[4].李宏毅 (2016)。Machine Learning。
(http://speech.ee.ntu.edu.tw/~tlkagk/courses_ML16.html )。
[5].斎藤康毅 (2017)。Deep Learning – 用Python進行深度學習的基礎理論實作。碁峰資訊股份有限公司。
[6].Nikhil Buduma (2018)。Deeping Learning 深度學習基礎 – 設計下一代人工智慧演算法。碁峰資訊股份有限公司。
[7].Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., and Fei-Fei, L. (2009). ImageNet: A Large-Scale Hierarchical Image Database. In CVPR09.
[8].Duchi, J., Hazan, E., and Singer, Y. (2011). Adaptive subgradient methods for online learning and stochastic optimization. Journal of Machine Learning Research,2121-2159.
[9].Glorot, X., Bordes, A., and Bengio, Y. (2011a). Deep sparse rectifier neural networks. In AISTATS’2011 .
[10].Goodfellow, I. J., Bengio, Y., and Courville, A. (2016). Deep Learning . https://www.deeplearningbook.org.
[11].He, K., Zhang, X., Ren, S., and Sun, J. (2015). Deep Residual Learning for Image Recognition. arXiv:1512.03385 .
[12].Kingma, D. and Ba, J. (2014). Adam: A method for stochastic optimization. arXiv:1412.6980 .
[13].Krizhevsky, A., Sutskever, I., and Hinton, G. (2012b). ImageNet classification with deep convolutional neural networks. In Advances in Neural Information Processing Systems 25(NIPS’2012).
[14].Qian, N. (1999). On the momentum term in gradient descent learning algorithms. Nerual networks: the official journal of the International Neural Network Society,12:145-151.
[15].Rosenblatt, F. (1958). The Perceptron: A probabilistic model for information storage and organization in the brain. Cornell Aeronautical Laboratory, Psychological Review, v65, No. 6, pp. 386–408.
[16].Ruder, S. (2017). An overview of gradient descent optimization algorithms. arXiv:1609.04747 .
[17].Simonyan, K. and Zisserman, A. (2014). Very deep convolutional networks for large-scale image recognition. arXiv:1409.1556
[18].Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., and Salakhutdinov, R. (2014). Dropout: A simple way to prevent neural networks from overfitting. Journal of Machine Learning Research, 15, 1929–1958.
[19].Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., and Rabinovich, A. (2014a). Going deeper with convolutions. Technical report, arXiv:1409.4842.
[20].Tieleman, T. and Hinton, G. ( 2012 ).Lecture 6.5- RMSProp:Divide the gradient by a running average of its recent magnitude. COURSERA: Neural Networks for Machine Learning. |