Loading...

Hebrid Generative Models of Social Networks

Mahdavi, Hamed | 2021

1052 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 53940 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Beigy, Hamid
  7. Abstract:
  8. With the advent of graph neural network networks, a new class of models has emerged that has a high ability to learn powerful representations. Also, there have been popular probabilistic latent variable models for representation learning and solving graph problems. Graph neural networks do not necessarily provide meaningful representation in their hidden layers and also do not have the ability to estimate uncertainty. Learning probabilistic models is usually a slow process and there is no specific way to add general features to these models. Therefore, recently, a combination of neural network models and probabilistic network models have been developed that can partially answer these problems. Therefore, recently, a combination of graph neural networks and probabilistic network models have been developed that can partially answer these problems. In this dissertation, several hybrid models of graph neural networks and probabilistic models that have the ability to extract network communities and meaningful representations of the network are presented. Also, unlike previous works, these models express communities explicitly as a distribution on the edges. In other words, in the proposed models, each vertex in a community has its own probability and the probability of the edge between two vertices in a community is proportional to the probability of these two vertices in that community. Also, as far as we know, in this research, for the first time, a model is presented that has used graph neural network in its decoder part. Finally, using empirical criteria, we discuss the advantages and disadvantages of the introduced models and show that these models have results close to the current developments and are able to learn and extract meaningful representations of the data
  9. Keywords:
  10. Graph Neural Network ; Social Networks ; Probabilistic Modeling ; Latent Variable Generative Models ; Deep Generative Modeling ; Representation Learning

 Digital Object List

 Bookmark

...see more