Graph Gurus Episode 19: Deep Learning Implemented by GSQL on a Native Parallel Graph Database

Recorded Wednesday, October 9 

Deep learning, as a class of artificial intelligence technology, has demonstrated its potential to produce results superior to human experts over a broad range of applications such as computer vision, speech recognition, autonomous driving, recommendation systems, and drug design. Unlike many machine learning methods that usually require the learning features provided to the model manually, deep learning utilizes multiple-layer artificial neural network to progressively extract higher level features from the raw input which allows it to automatically discover the features to be used for classification. Considering the ever growing size of the neural network model and the training data as well as the increasing complexity of the graphs to represent the neural networks, a graph database management system shows great advantages in scaling to large models and representing the neural network. 

In this Graph Gurus episode, we:

  • Review the basics of deep learning algorithm,
  • Introduce a classical classification problem: recognize a hand-written digit, 
  • Present a graph solution to build and train an artificial neural network for digit recognition using TigerGraph GraphStudio and GSQL,
  • Review a test dataset and GSQL queries for the solution.


  • Changran Liu, Solution Architect