What is Graphical Data
A Graph (V, E) is the type of data structure that contains nodes and edges.
- A node can be a person, place, or subject.
- An edge define the relationship between two nodes. The edges can be directed and undirected based on directional dependencies.
How to Model Graphic Data
It is not easy to model and analyze graphical data due to its graph-based data structures – A graph is in a non-euclidean space.
– Graphs can be dynamic. There can be two visually different graphs, but they might have similar adjacency matrix representations.
– Graphs can be large in terms of size (i.e., number of nodes) and dimensionality (number of edges). It poses great challenges to understand and extract useful information.
A Data Science Perspective for Graphical Neural Network
Graph Neural Network (GNN) is a type of neural network to deal with graph data structure.
– GNN, based on Convolutional Neural Networks (CNNs), can be used to predict a class label of a graph structure. – GNN, based on Recurrence Neural Networks (RNNs), can be used for text classification where the graph graph structure is with respect to every word to be a node in a sentence.
There are some limitations of GNN.
– GNNs often have two to three layers.
– The dynamic nature of the graph making model training difficult.
– Scalability issues make the computation expensive