Loading...

Design of Optical Convolutional Neural Network for Image Classification

Sadeghzadeh Bahnamiri, Hoda | 2022

126 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 55554 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Koohi, Somayyeh
  7. Abstract:
  8. Convolutional neural networks (CNNs) are at the heart of several machine learning applications, while they suffer from computational complexity due to their large number of parameters and operations. Recently, all-optical implementation of the CNNs has achieved many attentions, however, the recently proposed optical architectures for CNNs cannot fully utilize the tremendous capabilities of optical processing, due to the required electro-optical conversions in-between successive layers. Therefore, in our first study, we proposed OP-AlexNet which has five convolutional layers and three fully connected layers. Array of 4f optical correlators is considered as the optical convolutional layer, saturable absorption is considered as the optical nonlinearity unit, and finally, pinhole and Gaussian filter masks perform optical average pooling and optical motion pooling operations, respectively. However, to facilitate optical implementation of CNNs in more than one layer, optical summation of channels’ output for each convolutional kernel should be provided to feed the optical nonlinearity units. Moreover, although blurring the transmitted image by passing through a low pass filter (i.e. a pinhole mask) simulates the average pooling operation, an efficient AlexNet architecture utilizes a max pooling layer. Therefore, either optical implementation of max pooling operation or an optical function with similar behavior as a max pooling unit should be provided to optimize the network classification accuracy.Considering the lack of multi-layer photonic CNN implementation, in our second study, we explore a fully-optical design for implementing successive convolutional layers in an optical CNN. As a proof of concept, and without loss of generality, we considered two successive optical layers in the proposed network, named as 2L-OPCNN, for comparative studies against electrical counterpart and single optical layer CNN. Our simulation results confirm nearly the same accuracies for classifying images of Kaggle Cats and Dogs challenge, CIFAR-10, and MNIST datasets, compared to the electrical counterpart, as well as improved accuracies compared to single optical layer CNN.The classification performance of all-optical CNNs is greatly influenced by components’ misalignment and translation of input images in the practical applications. In this manner, in our third study, we propose a free-space all-optical CNN (named Trans-ONN) which accurately classifies translated images in the horizontal, vertical, or diagonal directions. Trans-ONN takes advantages of an optical motion pooling layer which provides the translation invariance property by implementing different optical masks in the Fourier plane for classifying translated test images. The comparative studies confirm that the Trans-ONN achieves appropriate classification accuracy for translated images of Kaggle Cats and Dogs, CIFAR-10, and MNIST datasets
  9. Keywords:
  10. Optical Computing ; Images Classification ; Convolutional Neural Network ; Deep Learning ; Translation-Invariant Products ; Optical Correlator ; Optical Nonlinearity ; Optical Pooling

 Digital Object List

 Bookmark

No TOC