Susha Suresh

 

Date:  Tuesday, June 11th, 2019

Time:  01: 00 pm – 03:00 pm

Location: 3105, Lambton Tower

From social networks to biological networks, graphs are a natural way to represent a diverse set of real-world data. This work presents DGCP, a novel end-to-end Deep Graph Convolutional Neural Network model which captures the higher order latent features of weighted, labelled, undirected, attributed graphs of arbitrary size with a fixed structure. The architecture of DGCP is inspired by Convolutional Neural Networks and has a linear filter function that convolves over the fixed topological structure of a graph in order to learn local and global features. Convolution is followed by a statistically significant pooling layer which coarsens the graph while preserving the global structure of the original input graph.

 

On the other hand, advances in high throughput technologies for genetic interaction analysis have enabled machine learning research to acquire and extract knowledge from biological networks. DGCP is

applied on gene interaction networks with genomic mutation features as the attributes of the nodes. The performance of graph classification is compared with two state-of-the-art machine learning models with respect to accuracy. DGCP is anticipated to yield better results relative to the previously proposed models.

 

Internal Reader: Dr. Jianguo Lu

External Reader: Dr. Esam Abdel-Raheem

Advisor: Dr. Dan Wu and Dr. Luis Rueda

 

 

5113 Lambton Tower, 401 Sunset Avenue, Windsor ON, N9B 3P4, (519) 253-3000 Ext 3716, csgradinfo@uwindsor.ca