Defence of dissertation in the field of Computer Science, MSc Nguyen Tran Quang

Fundamental limits of machine learning for networked data
CS_defence_3 photo by Matti Ahlgren

Title of the dissertation is " Machine Learning for Networked Data"

Data in many applications, such as social networks, biological systems, and the Internet, has intrinsic network structure, and thus, is referred to as networked data. Given the labels of few data points, one can apply (semi-supervised) machine learning methods to predict the labels of unlabelled nodes by exploiting the network structure. However, in some applications, the network structure might not be provided and has to be learned in a data-driven fashion. While the learning methods for these problems are well-studied, guaranteeing the accuracy of the methods is a less explored topic.

In this dissertation, we provide sufficient conditions such that accurate learning is possible for the learning problems within networked data. In particular, we introduce the network compatibility conditions, the conditions on the location of labelled nodes concerning the network structure, which guarantees that network lasso methods can accurately learn the networked predictive model. Besides, we derive the bounds on the number of samples such that a conditional independence graph can be learned accurately.

Opponent: Professor Joakim Jalden, KTH Royal Institute of Technology, Sweden

Custos: Professor Alexander Jung, Aalto University School of Science, Department of Computer Science

Contact information: Nguyen Tran-Quang, Department of Computer Science, +358469229631, [email protected]

Electronic dissertationThe dissertation is publicly displayed 10 days before the defence at the noticeboard of Aalto University School of Science in Konemiehentie 2

  • Published:
  • Updated: