ETD

Digital archive of theses discussed at the University of Pisa

 

Thesis etd-01162018-191118


Thesis type
Tesi di laurea magistrale
Author
ERRICA, FEDERICO
URN
etd-01162018-191118
Thesis title
A generative approach for learning contexts in graphs
Department
INFORMATICA
Course of study
INFORMATICA
Supervisors
relatore Prof. Bacciu, Davide
relatore Prof. Micheli, Alessio
Keywords
  • machine learning
  • structured domain
  • incremental approach
Graduation session start date
02/03/2018
Availability
Full
Summary
We present a novel method for unsupervised learning in graphs. It relies on a constructive approach to spread contextual information, which works by stacking layers of probabilistic models. The model is exploited in a supervised setting using an SVM. We describe how structural information can be captured, and empirical evidence shows that our model performs well for tasks involving trees and network data.
File