Image Categorization Using Deep Learning

Behfar, Sajad | 2017

1596 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 49319 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Jafari Siavoshani, Mahdi; Rabiee, Hamid Reza
  7. Abstract:
  8. The representation of data influences the explanation factors of data variations. Thus,the success of learner algorithms depends on the data representation. Our main contribution in this thesis is learning of high level and abstract representation using deep structure. One of the fundamental examples of representation learning is the AutoEncoders. The auto-encoder is a rigid framework that doesn’t consider explanation factors in terms of statistical concepts. So, the auto-encoders can be re-interpreted by seeing the decoder as the statistical model of interest. The role of encoder is a mechanism for inference in the model described by the decoder. Our purpose is to design such model with convolutional layers. We obtain deep architecture by stacking the autoencoder model. Also, we assert that such model learns its hierarchy on the distribution mean of the encoder. It is clear, this model is more flexible against the other deterministic models. For instance, in involving the prior knowledge or image completion felds. We test the proposed model on MNIST data set. The generative model is able to generate samples like data set. Also, classifcation of representations learned by encoder results in a good accuracy
  9. Keywords:
  10. Deep Learning ; Hierarchical Structure ; Exponential Family ; Generative Process ; Variational Autoencoder ; Variational Bayesian Inference

 Digital Object List