top of page
Meta network where each node represent a network of some type.

Sub/graph Classification

Network classification is at a nascent stage, holding great potential. The network classification research is facilitated by an abundance of network datasets available today. 

Many complex structures represented by networks exist in larger contexts, for example: text documents and the semantic web, terrorist cells and online social networks, metabolic pathways and metabolite correlation networks, software and shared libraries, and many more. All these structures can be defined as subgraphs within the context of a larger network. 

In this project we aim to solve the general problem of identifying sets of vertices whose interrelationships match a given class of graphs or subgraphs. In particular, we provide hypotheses of yet undiscovered metabolic pathways based on the analysis of metabolite correlation networks. The general methodology includes three types of algorithms: embedding: expert-based and unsupervised feature generation methods for subgraphs, classification: statistical models suitable for classifying subgraphs, and search methods for pinpointing subgraphs that best fit a given model. 

Subgraph embedding and classification, as investigated in this project, is a new unique approach that enables the study of complex discrete structures within larger contexts. Tools and methods developed in this project are of broad interest in multiple research disciplines. Pathway hypotheses produced during the research may be utilized by biologists during the study of unmapped metabolites.

Early Detection Alert and Response to eThreats (eDare): Project

Sub/graph Classification

References and Links to Papers

Early Detection Alert and Response to eThreats (eDare): CV

Michael Fire, Lena Tenenboim, Ofrit Lesser, Rami Puzis, Lior Rokach, Yuval Elovici, "Link Prediction in Social Networks using Computationally Efficient Topological Features" ,IEEE Third International Conference on Privacy, Security, Risk and Trust (PASSAT) and IEEE Third International Confernece on Social Computing (SocialCom) (2011) . [google]

Rami Puzis, Michael Firei, Yuval Elovici, "Link prediction in highly fractional ‎data ‎sets" ,Handbook of Computational Approaches to Counterterrorism ,V.S. Subrahmanian ,Springer (2012) . [google]

Michael Fire, Lena Tenenboim-Chekina, Ofrit Lesser, Rami Puzis, Lior Rokach, Yuval Elovici, "Computationally Efficient Link Prediction in Variety of Social Networks" ,ACM Transactions on Intelligent Systems and Technology ,5 (1): (2013) . [google]

Rami Puzis, Yedidya Bar-Zev, Arik Vartanian, "Graph Classification using Information-Gain Feature Ranking" NetSci, Berkeley, CA, US, 2014.

[abstract, poster]

Sukrit Gupta, Konstantin Kilimnik, Rami Puzis, "Comparative Network Analysis Using Kroneker Graphs" ,CompleNet (2016) ,Dijon, France. [google]

Rami Puzis, Zion Sofer, Dvir Cohen, Matan Hugi, "Embedding-Centrality: generic centrality computation using neural networks" ,CompleNet (2018) . [google]

David Toubiana, Rami Puzis, Lingling Wen, Noga Sikron, Aaron Fait, Assylay Kurmanbayeva, Aigerim Soltik, Moshe Sagi, Maria del Mar Rubio Wilhelmi, Nir Sade, Eduardo Blumwald and Yuval Elovici, "Correlation-based network analysis combined with machine learning techniques – a novel method for metabolic pathway detection", working paper.

bottom of page